Biomechanics of Vasc Soft Tissue

CISM COURSES AND LECTURES
Series Editors:
The Rectors
Manuel Garcia Velarde- Madrid
Mahir Sayir - Zurich
Wilhelm Schneider - Wien
The Secretary General

Bernhard Schrefter - Padua
Former Secretary General

Giovanni Bianchi - Milan
Executive Editor
Carlo Tasso- Udine
The series presents lecture notes, monographs, edited works and

proceedings in the field of Mechanics, Engineering, Computer Science
and Applied Mathematics.
Purpose of the series is to make known in the international scientific
and technical community results obtained in some of the activities
organized by CISM, the International Centre for Mechanical Sciences.
INTERNATIONAL CENTRE FOR MECHANICAL SCIENCES
COURSES AND LECTURES- No. 441
BIOMECHANICS OF SOFT TISSUE

IN CARDIOVASCULAR SYSTEMS
EDITED BY
GERHARD A. HOLZAPFEL
GRAZ UNIVERSITY OF TECHNOLOGY
RAYW.OGDEN
UNIVERSITY OF GLASGOW
t Springer-Verlag Wien GmbH

This volume contains 148 illustrations
This work is subject to copyright.

All rights are reserved,
whether the whole or part of the material is concerned
specifically those of translation, reprinting, re-use of illustrations,
broadcasting, reproduction by photocopying machine
or similar means, and storage in data banks.
ISBN 978-3-211-00455-5 ISBN 978-3-7091-2736-0 (eBook)

DOI 10.1007/978-3-7091-2736-0
© 2003 by Springer-Verlag Wien
Originally published by CISM, Udine in 2003.
SPIN 10911716
In order to make this volume available as economically and as

rapidly as possible the authors' typescripts have been
reproduced in their original forms. This method unfortunately
has its typographical limitations but it is hoped that they in no
way distract the reader.
PREFACE
This volume consists of the Lecture Notes for an Advanced School on "Biomechanics of Soft
Tissue" held at the International Centre for Mechanical Sciences (CISM) in Udine, Italy, during
the period September 10-14, 2001. The course was presented by 6 lecturers, 3 from Europe, 2
from the USA and 1 from Japan and was attended by nearly 70 participants from 20 countries.
Biomechanics in general, and the biomechanics of soft tissue in particular, did not really
become a clearly defined and active research field until the mid-1960s. It is suggested that there
were three major events in history necessary before the modern biomechanics of soft tissue could
evolve:
(i) the development of the nonlinear field theory of mechanics during the 1950s and
1960s;
(ii) the development of the finite element method in the 1950s and subsequently;
(iii) the rapid advances in computer technology.
Analyzing the mechanical response of any soft tissue presents many fascinating challenges
to various disciplines. Today's soft tissue mechanics is a distinct multidisciplinary field of rese-
arch involving disciplines such as experimental and applied mechanics, mathematics, biology,
engineering and clinical medicine. Interdisciplinary and multidisciplinary teams provide the key
to answering questions concerning the many diverse forms of biological soft tissues. What all
soft tissues have in common is that they are composed of cells (the fundamental structural and
functional units) plus other material such as the extracellular matrix (which serves several fun-
ctions in a strong symbiotic relationship with the cells). They are anisotropic, non-homogeneous
and pre-strained, exhibit highly nonlinear and time-dependent responses under different loading
conditions, undergo large deformations, and cannot always be treated within a purely mechani-
calframework. The study of both the biological and mechanical characteristics of tissue consti-
tuents is important for contributing a deeper understanding of soft tissue structure and function
and, in particular, the underlying mechanobiology. Of course, it is the complexity and the neces-
sity for multidisciplinary approaches that makes the analysis of soft tissue so challenging.
The focus in this volume is a review of our current knowledge of the behaviour of soft tis-
sues in the cardiovascular system under mechanical loads, and the importance of constitutive
laws in understanding the underlying mechanics is highlighted. There are many important pro-
blems found in health, rehabilitation, disease, injury, and clinical intervention that require an
understanding of the biomechanics of soft tissue. In particular, the clinical importance of this
understanding, which should be based on physiology, cell biology, physical and computational
models, and the solution of boundary and initial value problems, has never been greater.
One important example is balloon angioplasty, which is a well-established interventional
procedure aimed at reducing the severity of atherosclerotic stenoses. Balloon angioplasty is
the most frequently used therapeutical intervention world wide and attracts great and steadily
growing medical, economic and scientific interest. A total of 1,069,000 angioplasty procedures
(including 601,000 coronary angioplasty procedures) were performed in 1999 in the United
States alone. From 1987 to 1999 the number of procedures increased 285 percent and the
number of patients increased 286 percent 1• The knowledge that balloon angioplasty is a mecha-
1 American Heart Association. 2002 Heart and Stroke Statistical Update. American Heart Association, Dallas, Texas, 2001.
nical solution for a clinical problem implies the necessity for a detailed understanding of the
biomechanics and mechanobiology of the types of soft tissues and calcifications involved. It is
exactly this understanding that is needed to improve such procedures, which often fail due to
restenosis (about 40% within six months at the coronary site).
The lectures in the course aimed to present a state-of-the-art overview of the mechanics of
soft tissues, with particular reference to arteries and the cardiovascular system, but also inclu-
ding information on tendons, ligaments and other biological soft tissues of current research inte-
rest in biomechanics. The lectures included (i) essential background on the histological structure
of soft tissues, important knowledge for developing both microscopic and macroscopic material
models in order to understand the underlying physiological functions, (ii) details of experimental
techniques and the results of experiments, which provided information about the highly nonli-
near mechanical response of various soft tissues (and their component cells) under different
boundary loads, which are invariably spatially non-uniform and time-varying, (iii) continuum
mechanical modelling, which developed the nonlinear elasticity and viscoelasticity theory back-
ground, with particular reference to anisotropic response associated with fibre reinforcement,
(iv) computational perspectives, a particular emphasis being on the nonlinear behaviour, and
the application of finite element methods to the simulation of the biomechanics of intracranial
saccular aneurysms and the heart, and to clinical procedures such as balloon angioplasty and
stenting, (v) the topics of tissue remodelling and growth (important for, e.g., wound healing,
adaptation to arterial hypertension, aneurysm development, morphogenesis etc.), together with
aspects of arterial grafting and analysis of the influence of residual stress, ageing and degene-
ration on the mechanical characteristics.
We have pleasure in thanking our colleagues, Jay D. Humphrey, Kozabura Hayashi, Andrew
D. McCulloch, Alexander Rachev for presenting their lectures and for preparing and contribu-
ting chapters to this volume. We are also grateful for the participants, who contributed both
short talks on their own work and to lively discussions. Special thanks are due to F. Holzmann
for his contribution to the editing of parts of the manuscript. We thank particularly Professor
S. Kaliszky, Rector of CISM, and the staff of CISM for their assistance and hospitality, and Pro-
fessor C. Tasso, Executive Editor of CISM, for his encouragement to publish this collection of
lecture notes.
G.A. Holzapfel
R.W Ogden
CONTENTS
Page
Preface
The Cardiovascular System -Anatomy, Physiology and Cell Biology

by Jay D. Humphrey and Andrew D. McCulloch ................................................ I
Mechanical Properties of Soft Tissues and Arterial Walls

by Kozaburo Hayashi ........................................................................................ 15
Nonlinear Elasticity, Anisotropy, Material Stability and Residual Stresses in

Soft Tissue
by Ray W Ogden ............................................................................................... 65
Structural and Numerical Models for the (Visco )elastic Response of Arterial
Walls with Residual Stresses
by Gerhard A. Holzapfel ................................................................................ 109
Intracranial Saccular Aneurysms

by Jay D. Humphrey ........................................................................................ 185
Remodeling of Arteries in Response to Changes in their Mechanical Environ-

ment
by Alexander Rachev ....................................................................................... 221
Computational Methods for Soft Tissue Biomechanics

by Taras P. Usyk and Andrew D. McCulloch ................................................. 273
The Cardiovascular System - Anatomy, Physiology
and Cell Biology
J.D. Humphreyl and A.D. McCulloch 2

1Biomedical Engineering, Texas A&M University, College Station, TX77843-3120, USA
E-mail: jhumphrey@tamu.edu, Home Page: http://biomed.tamu.edu/faculty/humphrey/
2 Department of Bioengineering, University of California, San Diego, La Jolla, CA92093-0412, USA
E-mail: mcculloc@be-research.ucsd.edu, Home Page: http://cmrg.ucsd.edu/
Abstract. Biomechanics aims to explain the mechanics of life and living. From molecules
to organisms, everything must obey the laws of mechanics- Y.C. Fung. The primary func-
tion of the cardiovascular system is mass transport, that is, the transport of oxygen, carbon
dioxide, nutrients, waste products, hormones, etc., within the body. This system consists
primarily of the heart, which serves as the pump, the blood, which serves as the conducting
medium, and the vasculature, which serves as the conduit through which the blood flows.
Closely related systems are the cardio-pulmonary and reno-vascular systems. The primary
organs of the pulmonary system are the lungs, which facilitate the exchange of oxygen and
carbon dioxide between the blood and external environment, whereas the primary organs of
the renal system are the kidneys, which serve as filters to remove waste products from the
blood. Although studying the pulmonary and renal systems is important, interesting, and
challenging, we shall focus on the heart and vasculature in this chapter. Moreover, within
the vasculature there are numerous 'special' circulations, including the cerebral, coronary,
pulmonary, and fetal. Our emphasis will be on general characteristics of the vasculature and
heart, however.
1 The Heart
1.1 Basic Anatomy
The heart consists of four chambers, the left and right atria and left and right ventricles, the latter
two being separated by the interventricular septum. The atria receive blood from the body (the
left from the lungs, the right from the remainder) and the ventricles pump blood to the body (the
right to the lungs, the left to the remainder). One way flow is maintained within the heart by four
valves, the pulmonary, aortic, mitral, and tricuspid. The latter two valves separate the atria and
ventricles; they consist of two and three leaflets, respectively, and they are stabilized in part by
thin collagenous fibers (called chordae tendineae) that connect them to finger-like projections
of muscle within the ventricles, the papillary muscles. In addition to the distinct papillary mus-
cles, the interior surface of the heart is characterized by many trabeculae, or muscular ridges.
The blood supply to the wall of the heart originates via three primary vessels: the left anterior
descending artery (LAD), left circumflex artery (CIRC), and right coronary artery (RCA). In
general, the left coronaries supply the left atrium and the anterior and lateral portions of the left
ventricle (LV) and the right coronary artery supplies the right atrium and right ventricular free
wall. The posterior and inferior portions of the ventricles may receive blood from either the left
2 J.D. Humphrey and A.D. McCulloch
or right coronary systems. Furthermore, the interventricular septum, which separates the left and
right ventricles, is typically supplied by branches off the LAD.
The veins in the heart are divided into two systems. A large superficial system, the largest
vessels of which parallel the main coronary arteries, drains most of the coronary blood into the
right atrium, and a smaller, deeper system (e.g., thebesian veins) drains blood directly into any
of the four chambers.
The wall of the heart consists of three distinct layers: an inner layer called the endocardium, a
middle layer called the myocardium, and an outer layer called the epicardium. The endocardium
lines the inside of each of the cardiac chambers; it is a thin serous membrane(::::::: 100 mm thick)
consisting primarily of a 2-D plexus of collagen and elastin as well as a single cell layer of
endothelial cells that serve as a direct interface between the blood and the heart wall. The middle
layer, or myocardium, is the parenchymal (i.e., functional) tissue that endows the heart with its
ability to pump blood. The myocardium consists primarily of myocytes that are arranged into
locally parallel muscle fibers that are embedded in an extracellular matrix consisting largely of
types I and III collagen (collagen is the most abundant protein in the body). The orientations
of the muscle fibers change with position in the wall; in the equatorial region, for example,
the predominant muscle fiber direction changes from about -65° in the sub-epicardial region
to nearly oo in the mid-wall region to about 65° in the sub-endocardial region, all relative to
the circumferential direction. This transmural splay of fiber directions causes the heart to twist
during the cardiac cycle.
The outermost layer, or epicardium, is also a thin(::::::: 100 mm) serous layer, consisting largely
of a 2-D plexus of collagen and some elastic fibers. Noting that the epicardium is also called the
visceral pericardium, the heart is surrounded by yet another serous membrane, the pericardium
(or parietal pericardium). This membrane is thicker than the endocardium and epicardium, but it
also consists primarily of a 2-D plexus of collagen with some elastic fibers. The pericardium is at-
tached to the diaphragm inferiorly and the hilum superiorly, and it creates a small potential space
about the heart that is filled with a 'lubricating' pericardia] fluid (::::::: 25 ml in man). It is thought
that the pericardium serves as a type of cradle that limits gross motions of the heart, which is
merely suspended in the chest by its connection to the lungs and major blood vessels. These are
the major anatomical structures most commonly studied in the field of cardiac mechanics.
In this chapter, particular emphasis is given to the ventricular walls, which are the most
important structures with regard to the pumping action of the heart. From the perspective of
mechanics, the ventricles are thick-walled three-dimensional pressure vessels whose thickness
and curvatures vary regionally and temporally. In the normal heart, the ventricular walls are
thickest at the equator and the base of the left ventricle and thinnest at the left ventricular apex and
right ventricular free wall. Ventricular geometry has been best quantified in the canine heart. The
left ventricle, for example, is modeled reasonably well as a thick-walled ellipsoid of revolution
that is truncated at the base. The crescentic right ventricle wraps circumferentially around the left
ventricle about 180 degrees and extends longitudinally about two-thirds of the distance from the
base to the apex.
1.2 Histology and Cell Biology
The myocardium consists primarily of two cell types, cardiac myocytes and fibroblasts, plus an
abundant extracellular matrix (ECM). Although not capable of dividing, the cardiac myocyte is
The Cardiovascular System- Anatomy, Physiology and Cell Biology 3
biologically active: it is capable of synthesizing new proteins and thereby altering its size and
structure in response to changes in its environment (i.e., applied loads and the chemical milieu).
Indeed, myocytes may replace the bulk of their protein molecules every week or two via a bal-
ance between synthesis and degradation. Cardiac myocytes are typically I 0-20 J.Lm in diameter
and 80-125 J.Lm in length; they contain mainly myofibrils (1-2 ttm in diameter) within their
cytoplasm. Each myofibril consists of a string of tiny contractile units (::::; 2.2 J.Lm long) called
sarcomeres, and it is the end-to-end organization of the sarcomeres that gives cardiac muscle
its striated appearance. Each sarcomere consists of overlapping thin actin (5 nm in diameter)
and thick myosin (10-12 nm in diameter) filaments, an end view of which reveals a hexagonal
stacking arrangement.
The sliding filament theory of 1954 suggests that the myosin has many transverse arms, or
cross-bridges, that are about 13 nm long and connect to the actin. Contraction is initiated by the
release of ca++, which is sequestered in the surrounding sarcoplasmic reticulum, and results
in a smooth ratcheting action (at ::::; 15 J.Lm/sec) as the cross-bridges release, move forward, and
reattach thus shortening the sarcomere. The release of ca++ from the sarcoplasmic reticulum is
induced by an action potential that spreads from the cell membrane into the cell via T-tubules.
Note, too, that the sarcomere appears to be endowed with an elastic component- the proteins
titin and nebulin, which associate with the actin and myosin- that augments relaxation.
Conduction of electrical signals, and thus action potentials, from cell to cell is facilitated by
end-to-end and side-to-side connections (gap junctions) between myocytes. In particular, my-
ocytes are arranged in long columns, with the double membranes between opposing ends of the
cells being the intercalated disks. The latter attach the cells via desmosomes and serve to inter-
connect actin filaments from cell-to-cell. Cardiac myocytes are attached to an average of 11.3
neighbors, 5.3 on the sides and 6.0 on the ends.
The cardiac ECM consists primarily of the fibrillar collagens, types I (75-85%) and III (1 0-
20%), which are synthesized by the cardiac fibroblasts, the most abundant cell type in the heart.
Types IV and V collagen, which constitute the basement membrane of the myocytes, are also
found, however. Although once thought of as a relatively inert internal 'skeleton' for the my-
ocytes, it is now known that these collagens experience significant turnover, and thus endow the
heart with considerable adaptability. Albeit the major structural protein in connective tissues,
collagen only comprises 2-5% of the myocardium by dry weight, compared with the myocytes,
which make up 90%. The collagen matrix has a hierarchical organization that has been classi-
fied according to conventions established for skeletal muscle into endomysium, perimysium, and
epimysium. The percent of the type I collagen (relative to type III) in these three classifications
is 38%, 72%, and 84%, respectively.
The endomysium is associated with individual myocytes and includes a fine weave that sur-
rounds the cells and transverse structural connections (i.e., struts, 120-150 nm long) that connect
adjacent myocytes; the attachments localize near the Z-line of the sarcomere. Additional struts
appear to connect myocytes and capillaries. The primary purpose of the endomysium is proba-
bly to maintain registration between adjacent cells and between the myocytes and capillaries; its
predominately type III composition endows it with considerable deformability. The perimysium
groups cells together; it includes the collagen fibers that organize bundles of cells into laminar
sheets as well as large coiled fibers typically 1-3 J.Lm in diameter that are composed of smaller
collagen fibrils (40-50 nm ). These perimysial fibers may be the major structural elements of the
extracellular matrix, consistent with the dominance of type I over type III collagen.
Finally, the thick collagen sheaths of the epimysium that surround the myocardium form the
aforementioned epicardium and endocardium. It appears that collagen turnover is controlled by
many factors including circulating levels of angiotensin (ANG-II) and aldosterone as well as
their local production. Protease activity (e.g., matrix metalloproteinases or MMPs) most likely
plays an important role as well. In summary, the collagen network within the myocardium serves
many functions: it endows the heart with added structural integrity and by tethering myocytes
together it influences their contraction and relaxation; moreover, it protects the myocytes from
excessive stretching and serves to transfer actively generated force throughout the tissue. Indeed,
it is because of these functions that the 'passive stiffness' of the heart is an important determinant
of overall cardiac function. In addition to the collagens, the cardiac ECM also consists of some
elastic tissue and of course proteoglycans.
1.3 Pressure-Volume Behavior
The most basic mechanical parameters that describe the pumping action of the heart are the
generated blood pressure P and the volumetric flowrate Q. From the perspective of wall me-
chanics, the ventricular and pericardia! pressures are the most important boundary conditions.
The cardiac cycle is defined by the following. Ventricular filling, which immediately follows mi-
tral valve opening (MVO), is initially rapid because the ventricle produces a diastolic 'suction'
as the relaxing myocardium recoils elastically from its contracted systolic configuration below
the resting chamber volume. The later slow phase of ventricular filling (diastasis) is followed by
atrial contraction. The deceleration of the inflowing blood reverses the pressure gradient across
the valve leaflets and causes them to close (e.g., mitral valve closure MVC). This closure may
not be completely passive, however, because the atrial side of the mitral valve leaflets, unlike
the pulmonic and aortic valves, are cardiac in embryological origin, they have muscle and nerve
cells, and they are coupled electrically to the atrium.
Contraction of the ventricles is initiated by an electrical excitation, which is almost syn-
chronous (the duration of the QRS complex of the electrocardiogram (ECG) is only about 60
msec in the normal adult) and begins about 0.1 to 0.2 sec after atrial depolarization. Pressure
rises rapidly during the isovolumic contraction phase (about 50 msec in adult humans), and the
aortic valve opens (AVO) when the developed pressure exceeds the aortic pressure (afterload).
Most of the cardiac output is accomplished within the first quarter of the ejection phase, before
the pressure has peaked. The aortic valve closes (AVC) only 20-30 msec after the AVO, when
the ventricular pressure falls below the aortic pressure owing to the deceleration of the ejecting
blood.
The dichrotic notch, a characteristic feature of the aortic pressure waveform and a useful
marker of aortic valve closure, is caused in part by pulse wave reflections in the aorta. Because the
pulmonary artery pressure, against which the right ventricle pumps, is much lower than the aortic
pressure, the pulmonic valve opens before and closes after the aortic valve. Ventricular pressure
falls during isovolumic relaxation, and the cycle repeats. The pressure-volume curves for the right
ventricle are similar, but the pressures in the right ventricle and pulmonary artery are only about
a fifth of the corresponding pressures on the left side of the heart. The interventricular septum,
which separates the right and left ventricles, can transmit loads from one chamber to the other.
For example, an increase in right ventricular volume may increase the left ventricular pressure by
The Cardiovascular System- Anatomy, Physiology and Cell Biology 5
deforming the septum and thus reducing left ventricular volume. This direct interaction is most
significant during filling.
The phases of the cardiac cycle are customarily divided into systole and diastole. The end
of diastole (i.e., the start of systole) is generally defined as the time of mitral valve closure.
The end of systole is usually defined as the end of ejection, but it has been proposed to extend
systole until the onset of diastasis since there remains considerable myofilament interaction and
active tension during relaxation. This distinction is important from the point of view of cardiac
muscle mechanics; the myocardium is still active for much of diastole and may never be fully
relaxed at sufficiently high heart rates (over 150 beats per minute). Herein, however, we retain the
traditional definition of diastole, but consider the ventricular myocardium to behave 'passively'
only in the final slow-filling stage of diastole.
A useful approach for displaying changes in ventricular pressure and volume is the pressure-
volume loop. See Sagawa et al. (1988). The isovolumic phases of the cardiac cycle can be recog-
nized as the vertical segments of the loop, whereas the lower limb represents ventricular filling
and the upper limb the ejection phase. The difference on the horizontal axis between the vertical
isovolumic segments is the stroke volume, which expressed as a fraction of the end-diastolic vol-
ume is the ejection fraction. Changes in the filling pressure of the ventricle (preload) move the
end-diastolic point along the so-called end-diastolic pressure-volume relation (EDPVR), which
represents the passive filling mechanics of the chamber as determined primarily by the thick-
walled geometry and nonlinear elastic behavior of the resting ventricular wall. Alternatively, if
the afterload seen by the left ventricle is increased, stroke volume decreases in a predictable man-
ner. The locus of end-ejection points (AVC) forms the so-called end-systolic pressure-volume re-
lation (ESPVR), which is approximately linear over a variety of conditions and also largely inde-
pendent of the ventricular loading history. Hence, the ESPVR is almost the same for isovolumic
beats as for ejecting beats, although consistent effects of ejection history have been well char-
acterized. Sagawa et al. also suggested that connecting pressure-volume points at corresponding
times in the cardiac cycle also results in a relatively linear relationship throughout systole with
the intercept on the volume axis V remaining nearly constant. This leads to the global approxi-
mation that at any time during systole the ventricular volume v (t) is simply proportional to the
instantaneous pressure P(t) through a so-called time-varying elastance E(t):
P(t) = E(t)[v(t) - V], (1)
where v(t) and V are the current and reference volume, respectively. The maximum elastance
Emax, the slope of the ESPVR, has achieved considerable attention as an index of cardiac con-
tractility that is independent of ventricular loading. As the inotropic state of the myocardium
increases, as, for example, with the infusion of catecholamines, Emax increases; conversely, with
a negative inotropic effect, such as a reduction in coronary artery pressure, it decreases.
The area of the ventricular pressure-volume loop equals the work I; performed by the my-
1
ocardium on the ejecting blood:
ESV
I;= P(t)dv, (2)
EDV
where ESV stands for end-systolic volume and EDV for end-diastolic volume. Plotting this stroke
work against a suitable measure of preload gives a ventricular function curve that illustrates the
single most important intrinsic mechanical property of the heart as a pump. In 1914, Patterson
and Starling performed detailed experiments on a canine heart-lung preparation, and Starling
summarized their results with his famous 'Law of the Heart', which states that the work of the
heart increases with ventricular filling. The so-called Frank-Starling mechanism is now well
recognized to be an intrinsic mechanical property of cardiac muscle.
Stroke work is closely related to cardiac energetics. Because contraction of the myocardium
is fueled by adenosine triphosphate (ATP), 90-95% of which is normally produced by oxidative
phosphorylation, consumption of energy by the heart is often studied in terms of its oxygen con-
sumption, Y0 2 (ml 0 2 g- 1 beat- 1 ). Given that energy is also expended during non-working con-
tractions, one can define the pressure-volume area PYA (J g- 1 beaC 1 ) as the loop area (stroke
work) plus the end-systolic potential energy II (internal work), which is the area under the ES-
PYR left of the isovolumic relaxation line. That is,
PVA = E+Il. (3)
The PYA has a strong linear correlation with Y0 2 independent of ejection history. Typical values
for the dog heart are
vo2 = o.I2(PVA) + 2.0 x 10- 4. (4)
The intercept represents the sum of the oxygen consumption for basal metabolism and the energy
associated with activation of the contractile apparatus, which is altered by calcium and other
inotropic interventions (Sagawa eta!., 1988).
The primary determinants of the end-diastolic pressure-volume relation (EDPYR) are the
material properties of resting myocardium, the chamber dimensions and wall thickness, and
the boundary conditions at the epicardium, endocardium, and valve annulus. The EDPYR has
been approximated by an exponential function, although a cubic polynomial also works well.
Consequently, the passive chamber stiffness dP / d V is approximately proportional to the filling
pressure. Important influences on the EDPYR include the extent of relaxation, ventricular inter-
action and pericardia! constraints, and engorgement of the coronary vasculature. The material
properties and boundary conditions associated with the septum are also important for they de-
termine how the septum deforms. Through septal interaction, the end-diastolic pressure-volume
relationship of the left ventricle may be affected directly by changes in the hemodynamic load-
ing conditions of the right ventricle. Of course, the ventricles also interact indirectly since the
output of the right ventricle is returned as the input to the left ventricle via the pulmonary circu-
lation. The pericardium provides a low friction mechanical enclosure for the beating heart that
constrains acute ventricular overextension and it contributes to direct ventricular interactions and
augments the mechanical coupling between the atria and ventricles. For more on general cardiac
physiology, see Opie ( 1998), and for further discussion of cardiac biomechanics, see McCulloch
( 1995).
2 The Vasculature
2.1 The Vascular Tree
The vasculature is generally divided into two sub-systems, pulmonic and systemic. The former
relates to blood flow to, in, and from the lungs; the latter relates to all other blood flow in the
body. The main systemic artery is the aorta, which consists of an ascending portion, the so-called
The Cardiovascular System -Anatomy, Physiology and Cell Biology 7
arch, and the descending thoracic and abdominal segments. Large diameter vessels arising from
the human aorta include the iliacs (which supply the legs), one subclavian artery, one common
carotid artery, and the brachio-cephalic trunk, which gives rise to a second subclavian artery
(they supply the arms) and a second common carotid artery (they supply the neck and head).
Some of the other major branches off the aorta are the coronaries, intercostals, mesenteries, and
renals. These are but a few of the many specialized arteries that are found throughout the body.
There are, for example, over 100,000 systemic arteries in the adult canine which supply some 3
million arterioles and 3 billion capillaries. The canine pulmonic system similarly includes over
1000 arteries, with millions of associated arterioles and billions of capillaries.
The aorta tapers in diameter from its beginning at the aortic valve to its end at the aorta-
iliac junction. The associated change in cross-sectional area a can be approximated by a =
a 0 exp(-kx/r 0 ), where a 0 is the area at the aortic valve, with r 0 the radius, x the distance
from the valve, and k a taper-factor that typically ranges from 0.01 to 0.05. In humans, for
example, the mean inner diameter and wall thickness of the aorta are ~ 2.5 em and 2.0 rom,
respectively. Although the branches that arise from most arteries have a smaller lumen (i.e.,
inner diameter) than that of the parent vessel, the total cross sectional area through which the
blood flows increases significantly with distance from the heart- nearly 10~15 fold from the
aorta to the small arteries, some 25 fold from the aorta to the arterioles, and over 100 fold from
the aorta to the capillaries. The large luminal area of the capillary bed and the overall decrease
in the mean forward velocity of the blood from the aorta to the capillaries (e.g., from 33 to 0.3
cm/s in the canine) both facilitate the transport processes within the capillaries.
2.2 Histology
The microstructure of the normal arterial wall varies with location along the vascular tree, age,
species, local adaptations, disease, etc.; thus one must focus on the particular vessel and condi-
tion of interest. Nonetheless, arteries can be categorized according to two general types: elastic
arteries, which include the aorta, main pulmonary artery, common carotids, and common iliacs,
and muscular arteries, which include the coronaries, cerebrals, femorals, and renals. That is,
elastic arteries tend to be larger diameter vessels located 'closer' to the heart whereas muscular
arteries are smaller diameter vessels located closer to the arterioles. Transitional arteries, as, for
example, the external carotids, exhibit some characteristics of the elastic and muscular types.
Regardless of type, all arteries consist of three layers: the tunica intima, tunica media, and
tunica adventitia. The intima is similar in most elastic and muscular arteries, typically consisting
of a monolayer of endothelial cells and an underlying thin ( ~ 80 nm) basal lamina. Exceptions,
however, include the aorta and coronary arteries in which the intima may contain a subendothelial
layer of connective tissue and axially oriented smooth muscle cells. Endothelial cells are usually
flat and elongated in the direction of blood flow, often about 0.2~0.5 p,m thick, 10~15 p,m wide,
and 25~50 p,m long; exceptions occur near bifurcations wherein the blood flow is complex and
the cells are often polygonal in shape. Endothelial cells are inter-connected by both tight occlud-
ing junctions, which regulate the transport of substances across the endothelium, and in-plane
gap junctions, which allow cell-to-cell communication via the transport of ions and metabolites.
Endothelial cells may communicate directly with underlying smooth muscle cells via short, blunt
processes that extend through the basal lamina and into the media. The basal lamina (which is
sometimes referred to as the basement membrane) consists largely of net-like type IV colla-
gen, the adhesion molecules laminin 1 and fibronectin, and some proteoglycans; it provides some
structural support to the arterial wall, but acts primarily as an adherent meshwork on which the
endothelial cells can grow.
The internal elastic lamina separates the intima and media, but is often considered to be part
of the latter. A little thicker in muscular arteries, this lamina is essentially a fenestrated 'sheet'
of elastin that allows the transport of H 2 0, nutrients, electrolytes, etc. across the wall as well
as direct transmural cell-to-cell communication. The media contains smooth muscle cells that
are embedded in an extracellular plexus of elastin and collagen (primarily types I, III, and V) as
well as an aqueous ground substance matrix containing proteoglycans. Vascular smooth muscle
cells are spindle-shaped: they are typically 100 pm long and about 5 pm in diameter, except near
the nucleus where they are slightly thicker. Given this shape, they are often laid down such that
the thicker portion of one cell is juxtaposed to the thin ends of the neighboring cells. Smooth
muscle cells are covered by a thin (40-80 nm) basement-type membrane, hence 12-50% of the
volume of vascular 'smooth muscle' is due to this investing connective tissue. Like endothelial
cells, smooth muscle cells communicate, in part, via gap junctions. Their intracellular myofibrils
are typically oriented along the long axis of the cell, and cell-to-cell force transmission is ac-
complished via thin (reticular) collagen fibers that connect the membranous sheaths on each cell.
Although the orientation and distribution of the medial constituents vary with species and loca-
tion along the vascular tree, vascular smooth muscle tends to be oriented helically, albeit nearly
circumferentially in many vessels (Rhodin, 1979). This preferential orientation is expected, in
part, for the primary roles of the contraction of vascular smooth muscle are to modify the dis-
tensibility of the large arteries or to regulate the luminal diameter in medium and small arteries.
Consistent with their different roles in large and small vessels, the smooth muscle is packaged
somewhat differently in elastic and muscular arteries.
In elastic arteries, the medial smooth muscle is organized into 5-15 pm thick concentric 'lay-
ers' that are separated by thin (3 pm) fenestrated sheets of elastin. These elastin sheets are similar
to the internal elastic lamina, with the fenestra allowing radial transport of metabolites across the
wall, direct transmural cell-to-cell communication, and radially oriented connective tissue to tie
the wall together three-dimensionally. The outermost sheet of elastin is called the external elas-
tic lamina; it separates the media and adventitia but is often considered to belong to the former.
There may be as many as 40 to 70 concentric layers of smooth muscle in a thick elastic artery
such as the human aorta. Indeed, in 1985 Clark and Glagov (see Humphrey, 2002) suggested that
the alternating layers of smooth muscle and elastic lamina in elastic arteries (specifically, in ab-
dominal and thoracic aorta) represent a discrete structural and functional unit, the musculo-elastic
fascicle. Glagov and colleagues showed that the thickness of these units is nearly independent
of radial location in the wall, but their number increases with increased diameter of the ves-
sel (measured at the mean blood pressure). For example, a 1.2 mm diameter mouse aorta has 5
musculo-elastic units whereas a 23 mm diameter pig aorta has 72 units. A corollary, therefore, is
that the thickness of the aorta increases:::::: 0.05 mm for every 1 mm increase in diameter. More-
over, they showed that whereas total wall tension is significantly higher in the larger diameter
1 Laminin is one of the first extracellular matrix proteins synthesized in the embryo; it has numerous
functional domains that bind to type IV collagen, heparin sulphate, and laminin receptors on cells, etc.
(Alberts eta!., 1994)
vessels (e.g., 203 N/m in the pig versus 7.82 N/m in the mouse), the mean tension per lamellar
unit is nearly independent of aortic diameter (e.g., about 2 ± 0.4 N/m).
Here, tension T was calculated using Laplace's relation for a cylinder, that is T = Piri
where Pi is luminal pressure and r i the current inner radius. Glagov et al. also showed that
the elastic lamina become straight axially at the in vivo length and straight circumferentially at
pressures 2: 60-80 mmHg; this pressure dependent straightening occurs uniformly across the
wall, which is likely related to the existence of residual stress in the arterial wall as discussed
below. Collectively, these findings suggest that smooth muscle is packaged as units in elastic
arteries, the number of which ensures an optimal distribution of stress or strain throughout the
normal wall. This plays an important role in growth and remodeling, as discussed in the chapter
by Rachev in this volume.
In muscular arteries, on the other hand, the smooth muscle appears as a 'single', thick layer
that is bounded by a thick internal and less marked external elastic lamina. Exceptions, again,
are the cerebral arteries which do not have an external elastic lamina and indeed have few elastic
fibers in the media. Nonetheless, the smooth muscle in muscular arteries is embedded in a loose
connective tissue matrix and arranged as a sequence of concentric layers of cells; the number of
layers can reach 25-35 in larger vessels. The connective tissue augments the structural integrity
of the wall, including its ability to generate force, and acts as a scaffolding on which the cells can
adhere or move.
Finally, the adventitia, or outermost layer of the wall, consists primarily of a dense network
of type I collagen fibers with admixed elastin, nerves, fibroblasts, and the vasa vasorum. The
adventitial collagen fibers tend to have an axial orientation in most arteries and they are undulated
slightly in the basal state. Although the adventitia comprises only:=::::! 10% and 50% of the arterial
wall in elastic and muscular arteries, respectively, it is thought to limit acute over-distension in all
vessels. That is, the collagenous adventitia may serve primarily as a protective sheath, similar to
the epicardium of the heart. Nonetheless, the presence of nerves within the adventitia also allows
innervation of smooth muscle in the outer media via the diffusion of neurotransmitters, primarily
norepinephrine (NE) and acetylcholine (ACh). The fibroblasts are responsible for regulating the
connective tissue, particularly the type I collagen which experiences considerableturnover, again
similar to that in the heart. The vasa vasorum is an intramural network of arterioles, capillaries,
and venules that serves the outer portion of the wall in arteries that are too thick for transport of
Oz, COz, nutrients, metabolites, etc. from the intimal surface to suffice; it is generally accepted,
for example, that a vasa vasorum is needed if the number of concentric elastic lamella exceeds
29 in an elastic artery. In many vessels, the adventitia is contiguous with the perivascular tissue,
which often provides additional structural support. Exceptions, again, are the cerebral arteries,
which are sometimes surrounded mainly by cerebral spinal fluid. In summary, however, it is the
distributions, orientations, and inter-connections of the intramural constituents that give rise to
the mechanical properties of the arterial wall.
Further discussion of the histology of the arterial wall is contained in the chapter by Holzapfel
in this volume.
2.3 Physiology and Cell Biology
The aforementioned classification of arteries as elastic or muscular actually arises more from
considerations of function than structure. To appreciate this, consider the following. The heart
is a pulsatile pump that ejects blood into the vasculature only during its contractile phase (i.e.,
systole, which constitutes about 1/3 of the cardiac cycle). Whereas contraction of the heart results
in forward flow of most of the ejected blood, some of the blood is 'stored' in the large arteries as
they distend under high pressure. When the blood pressure drops during diastole, these distended
vessels 'recoil elastically' and thereby augment the flow of blood by supplying a second pressure
pulse. Indeed, coronary blood flow in the left heart is due primarily to recoil of the distended
aorta. Moreover, augmentation of blood flow via the recoil oflarge arteries helps convert it from
a pulsatile to a nearly steady flow in the capillaries, which in tum aids the requisite transport
processes. It is, of course, the large amount of intramural elastin and type III collagen in the media
that gives rise to this important elastic behavior of the large arteries, and hence the terminology.
In contrast, the smaller arteries (and especially arterioles) play a key role in regulating local
blood flow via vasoconstriction or vasodilation, that is by muscular activity. Note, therefore, that
arterial smooth muscle is partially contracted in its homeostatic or basal state -this is referred
to as basal tone. Further contraction or alternatively relaxation of the abundant smooth muscle in
these vessels thus results in a decreased or increased lumen, which in tum controls the resistance
to blood flow. It is in this way that blood can be routed to regions having an increased need for
oxygen or nutrient exchanges (e.g., skeletal muscles during exercise or the digestive system fol-
lowing eating) or away from regions where there is an injury to the vasculature (e.g., to minimize
bleeding). Likewise, because of their ability to control blood flow, muscular arteries and arteri-
oles play a key role in the thermal regulation within the body by routing blood towards or away
from the skin. Muscular arteries are thus those vessels which perform their primary function
via smooth muscle activity. Quantifying the biomechanics of the contractile activity of vascular
smooth muscle is vital, therefore, to understanding even the basics of vascular physiology.
As we have seen, the three primary cells within the arterial wall are the endothelial cells of
the intima, the smooth muscle cells of the media, and the fibroblasts of the adventitia. Recent
advances in molecular and cell biology have been spectacular, and such information is vital
to the development and testing of comprehensive theories of biomechanics. Here, however, we
consider only a few of the salient aspects of the cell biology. For more information, see Alberts
et al. ( 1994) and related works.
The luminal surface of the endothelial cell contains heparan-sulfate, a glycosaminoglycan
that is a potent anticoagulant. This property of the endothelium is in contrast to, for example, the
highly thrombogenic character of the intramural type I and III collagen (note: a thrombus is an
aggregation of blood factors, primarily platelets and fibrin, with an entrapment of other cellular
elements, which together can obstruct the lumen of a vessel). Clearly, then, an important role of
the endothelium is to act as a smooth, non-thrombogenic lining that separates the wall contents
from the flowing blood, and thereby augments the latter. The endothelium allows transport of sub-
stances to and from the bloodstream, however, hence it must be considered as a selective barrier.
Notwithstanding the importance of these functions of an intact endothelium, it was long thought
that these were its only two roles. It is now known, however, that the endothelium is very active
biologically: it can regulate coagulative processes (e.g., by expressing tissue plasminogen acti-
vator, t-PA, or its inhibitors, von Willebrand factor, etc.), it can synthesize vasoactive substances
(e.g., nitric oxide (NO), prostacyclin (PGI 2 ), endothelin-1 (ET-1 ), angiotensin-II (AN G-Il), etc.),
it can produce growth-regulatory molecules (e.g., vascular endothelial growth factor (VEGF),
platelet-derived growth factor (PDGF), fibroblast growth factor (FGF), etc.), it can synthesize
connective tissue (e.g., type IV collagen and proteoglycans), and it can modify plasma lipopro-
teins for transport into the arterial wall. Moreover, many blood-borne vasoactive substances, such
as acetylcholine, histamine, thrombin, etc., affect the arterial wall only in the presence of an in-
tact endothelium, thus it also serves as a chemo-transducer. Hence, albeit a small fraction of the
arterial wall, the endothelium plays a major role in vascular mechanics. Conversely, many bio-
logical functions of the endothelium are signaled mechanically, as, for example, via changes in
local blood flow and the associated wall shear stress: increased flow down-regulates the synthesis
of the vasoconstrictor ET-1 and upregulates that of the vasodilators NO and PGh, thus result-
ing in flow-mediated vasodilation. Endothelial cells also tend to elongate and align themselves
in the direction of the blood flow, which thereby reorients their actin microfilaments and alters
cytoskeletal properties.
Smooth muscle cells comprise about 25-60% of the arterial wall by dry weight (Milnor,
1990). They can increase in size (i.e., hypertrophy) or increase in number via cell division (e.g.,
via hyperplasia) depending on the stimulus. Structurally, smooth muscle cells contain a ca++
regulated actin-myosin contractile apparatus, albeit not arranged in sarcomeres as in striated
muscle, and they can possess a marked rough endoplasmic reticulum and well-developed Golgi
apparatus. The latter enables the synthesis of connective tissue, a capability that is very important
during development but also plays a (often detrimental) role in aging and different pathologies.
Note, too, that various signals are capable of changing the phenotype of the mature smooth
muscle cell from contractile to synthetic. Contraction of vascular smooth muscle occurs much
slower than that of striated muscle, typically beginning 5-100 ms after the initial stimulus and
taking on the order of 1-10 seconds to achieve its maximum level. Nevertheless, smooth muscles
can maintain maximum contraction at a steady level for much longer periods than striated muscle,
provided a stimulus persists, and with less energy expenditure (perhaps as little as 0.25-5% of
the energy required for comparable contraction of skeletal muscle). This ability is essential in
the maintenance of basal tone, which results from an integrated balance between the continuous
production of NO by the endothelium and NE by the neurotransmitters. Reasons for the different
function of vascular smooth muscle are not completely understood, but they undoubtedly result
from unique actin-myosin interactions. For example, vascular smooth muscle has a higher ratio
of actin to myosin (5:1) and longer myosin filaments(~ 2 pm).
Despite some differences, the general relations between the force, length, and velocity of
shortening are similar for smooth and striated muscle, both being able to generate 3 kg or more
of 'force' per cm 2 of muscle, that is an active muscle stress of~ 0.3 MPa. Studies on the overall
mechanics of vascular smooth muscle are often conducted in vitro using thin arterial rings that
are immersed in a thermally and chemically-controlled organ bath and subjected to uniaxial
extension. Like skeletal muscle, smooth muscle generates its maximum (isometric) force at a
unique length, often denoted La. That is, there is a length at which the muscle prefers to act,
it being able to generate less active force at lengths less than or greater than La. In contrast to
striated muscles, however, which develop their maximal active force at a length where passive
tension is absent, a significant passive force is required to achieve La in smooth muscle prior
to activation; this observation will prove important in constitutive modeling. Finally, note that
uniaxial tests also allow one to investigate easily dose-response curves for various pharmacologic
agents. Many similar studies are reported in the literature for various vasoactive agents, but the
key observation here is that both the concentration of the agonist and the length of the muscle
play key roles in the mechanics.
On the molecular level, intracellular free calcium ca++ is the primary determinant of con-
tractility. Cytoplasmic ca++ ranges from 100 nM in resting to 600-800 nM in fully contracted
smooth muscle. Thus, increases in intracellular ca++ associate with contraction and decreases
with relaxation. The two primary sources of intracellular ca++ are the sarcoplasmic reticulum
and the extracellular milieu, Ca ++ from which can enter the muscle through ion channels, pumps,
and exchangers in the cell membrane. As in striated muscle, smooth muscle can be activated via
changes in the potential of the cell membrane (the resting value of which is -40 to -60 m V): de-
polarization increases intracellular Ca ++, hence inducing contraction, whereas hyperpolarization
induces relaxation. Membrane potential can be changed in various ways, including electrically,
mechanically (thought to be stretch-induced), and chemically. For example, the neurotransmit-
ters ACh and NE act in part by changing the membrane potential. Moreover, since membrane
potential is governed largely by the ratio of intra- and extra-cellular potassium K+, a common
method of inducing smooth muscle contraction in tests on intact, excised arteries is by bathing
them in solutions of greater than 50-80 mM KCl (as revealed by dose-response studies). Many
other factors control smooth muscle contraction, including local concentrations of0 2 , C0 2 , ni-
tric oxide (NO), prostacyclin (PGI2), endothelin-1 (ET-1 ), etc. Specific effects of contraction on
the overall mechanics of the arterial wall are discussed in Section 3 below. Finally, fibroblasts
are important players in the maintenance of the adventitia, particularly via the synthesis of Type
I collagen.
3 Some Common Diseases
Cardiovascular disease is the leading cause of death in developed countries and it is projected
to become the leading cause of death worldwide in the near future. Although there are many
different diseases that merit biomechanical study, here we simply list a few.
Aneurysms are focal dilatations that result from a local weakening of a pressure distended or-
gan. Within the vasculature, the two most common forms are abdominal aortic aneurysms (AAA)
and intracranial saccular aneurysms. These two types of lesions have very different etiologies,
the former often related to atherosclerosis and the latter not; both generally involve proteolytic
breakdown of portions of the extracellular matrix, however. When possible, AAAs are typically
treated via the surgical replacement by a synthetic arterial graft; among many others, Albert
Einstein died of a ruptured AAA.
Atherosclerosis is characterized by a focal accumulation of lipids, extra cells and proteins,
calcium, and necrotic debris within the intimal layer of arteries. In most cases this accumulation
causes a narrowing of the lumen (i.e., a stenosis) which thereby compromises distal blood flow.
One of the primary complications, however, is that atherosclerotic plaques can rupture and then
clot - this can result in either the complete occlusion of the blood vessel or the shedding of a clot
which occludes a smaller, distal vessel. Consequences can include myocardial infarction (heart
attack) or stroke (brain attack).
Heart Failure is characterized by a marked decrease in cardiac output, increased venous
pressures, or both; this results in elevated ventricular pressures that overdistend the ventricles
and thereby diminish their ability to pump blood. Heart failure is generally classified according
to whether the right or left ventricle is affected. Left heart failure leads, for example, to pulmonary
edema and hence breathlessness - sometimes called congestive heart failure.
Hypertension is defined as the persistent elevation of blood pressure. Normal systemic pres-
sure is >::! 120/80 mmHg (systolic/diastolic) and systemic hypertension is generally defined as
pressures above 160/90 mmHg; pulmonary hypertension is generally defined as pressures above
30/12 mmHg. Hypertension may not have a known cause (called essential or idiopathic) or it
may result from other diseases (then called secondary hypertension), such as renal, endocrine, or
central nervous system diseases. The primary importance of hypertension is that it is a major risk
factor for other, potentially fatal, diseases, including aneurysms, end-stage renal disease, stroke,
sudden cardiac death, etc.
Myocardial Infarction, or heart attack, is defined as the death of myocardium (necrosis) due
to a lack of oxygen (ischemia). If one survives a heart attack, the necrotic tissue is generally
removed and replaced with a 'collagen patch', which in some cases may form an aneurysm. As
noted above, atherosclerosis is the most common cause of heart attacks.
Stroke, strictly speaking, is any sudden or severe attack, as, for example, a sun-stroke or heat-
stroke. Commonly, however, by the term stroke we imply death of a portion of the brain due to
the lack of oxygen, and thus it is sometimes referred to as a brain attack in analogy with a heart
attack. As noted above, strokes are often caused by the rupture and clotting of an atherosclerotic
plaque. In addition, however, strokes may also be caused by the shedding of an embolus from the
heart or a proximal vessel or by the rupture of an intracranial malformation or aneurysm.
Valvular disease often necessitates the surgical replacement of one of the heart valves; pros-
thetic heart valves are among the most successful implants in Cardiovascular Surgery.
4 Summary
Although thought by many to be the domain of physicians, surgeons, and biologists, the diag-
nosis and treatment of cardiovascular diseases and injuries also requires, in many cases, well
trained engineers and mathematicians. The purpose of this chapter was simply to provide a brief
introduction to some of the biological and medical terminology and issues. The serious investi-
gator must consult more complete sources, however. In particular, investment in a good medical
dictionary, a text on molecular and cell biology, a book on histology, and a book on cardiovas-
cular physiology is a must. In addition, of course, because biological and medical research is
occurring at such a rapid pace, it is essential to stay current with the archival literature - to this
end, review articles serve us well.
Acknowledgements
The authors acknowledge funding from the National Institutes of Health, National Science Foun-
dation, and the Whitaker Foundation.
References
Alberts, B., Bray, D., Lewis, J., Raff, M., Roberts, K. and Watson, J.D. (1994). Molecular Biology of the
Cell. New York: Garland Publishing.
Humphrey, J.D. (2002). Cardiovascular Solid Mechanics: Cells, Tissues, and Organs. New York: Springer-
Verlag.
McCulloch, A.D. (1995). Cardiac biomechanics. In Bronzino, J., ed., The Biomedical Engineering Hand-
book, Chapter 31, Section III. Boca Raton: CRC Press.
Milnor, W.R. (1990). Cardiovascular Physiology. Oxford: University Press.
Opie, L.H. (1998). The Heart: Physiology,from Cell to Circulation. New York: Lippincott-Raven.
Rhodin, J.A.G. (1979). Architecture of the vessel wall. In Berne, R.M., ed., Handbook of Physiology, Sec-
tion 2, Volume 2. Betesda: American Physiological Society.
Sagawa, K., Maughan, L., Suga, H. and Sunagawa, K. (1988). Cardiac Contraction and the Pressure-
Volume Relationship. Oxford: University Press.
Note: most definitions herein were taken from Dorland's Illustrated Medical Dictionary (1988)
W.B. Saunders Co., Philadelphia.
Mechanical Properties of Soft Tissues and Arterial Walls
Kozaburo Hayashi
Graduate School of Engineering Science, Osaka University

Biomechanics Laboratory, Division of Mechanical Science
Toyonaka, Osaka 560-8531, Japan
E-mail: hayashi@me.es.osaka-u.ac.jp
Home Page: http://www-biomech.me.es.osaka-u.ac.jplhayashiken-eng.html
Abstract. Mechanical properties of biological tissues are fundamental and prerequisite

for biomechanics. Basic mechanical properties, in particular those unique to biological
soft tissues, and their mathematical formulation are described for several tissue examples.
Then, the structure and composition of arterial walls are discussed along with the pressure-
diameter and stress-strain relations and their mathematical description. The effects of pul-
sation, smooth muscle contraction, arterial site and aging on the mechanical properties are
included in the discussion. Because of the importance of cellular mechanics in the physio-
logical function of tissues and organs and their diseases, the mechanical properties of cells
are also described together with several methodologies and techniques which have been
used for the determination of the properties. Biomechanics is very useful for analyzing the
pathogenesis of vascular diseases. Several examples of the application of biomechanics to
arterial diseases are therefore examined, including arterial wall elasticity in atherosclerosis
and hypertension, and the mechanical properties and vasospasm of cerebral arteries.
1 Mechanical Properties of Biological Soft Tissues

Biological tissues are roughly divided into: (i) hard tissues like bone and tooth, and (ii) soft
tissues such as skin, muscle, blood vessel, and lung. Hard tissues contain mineral, whereas soft
tissues do not. Because of this, they have very different mechanical properties. One of the major
differences in mechanical properties is that soft tissues are much more deformable than hard
tissues. Therefore, infinitesimal deformation theories that are applied to metals and hard plastics
cannot be used for soft tissues; instead, finite (large) deformation theories that are useful for
rubber elasticity are often used to describe the mechanical behavior of soft tissues.
This section deals with the basic mechanical properties of biological soft tissues and their
mathematical formulation, including several biomechanical features unique to soft tissues.
1.1 Test Methods

One of the most basic mechanical test methods is tensile testing, in which load and deformation
are measured while the force applied to a simple-shaped test specimen is gradually increased.
Creep and relaxation tests are sometimes performed to determine the viscoelastic properties of
soft tissues; specimen deformation is measured under a constant force in creep testing, whereas
in relaxation tests, change in load is determined under constant specimen length. In these tests,
stress (= load divided by undeformed specimen cross-sectional area) and strain (= change in
specimen length divided by reference length) are commonly used to represent the data.
16 K. Hayashi
200 30000 15
Nuchal Sole Intestinal

ligament tendon smooth
150 (Elastin) (Collagen) muscle
,-..,
ell
20000 10
~
'-'
h 100
"'"'
(!)
1-<
......
r:/1 10000
50
0 0.25 0.5 0 0.04 0.08 0.12 0 0.25 0.5

Strain E
Figure 1. Tensile properties of elastin-rich canine nuchal ligament, collagen-rich sole tendon, and intestinal
smooth muscle (Hasegawa and Azuma (1974)).
Inside the body (in vivo), tissues are not exposed to such a pure tensile force condition.
Therefore, multi-axial tests are often used to determine the behavior of tissues under more real-
istic mechanical conditions. For example, tri-axial tests are used for blood vessels, as described
in the next section, and bi-axial tests are applied to skin.
1.2 Basic Properties Unique to Soft Tissues
1. Inhomogeneous Structure
Biological soft tissues are composed mainly of cells and intercellular substances, the latter con-
sisting of connective tissues such as collagen and elastin, and ground substance (hydrophilic gel).
These components have different physical and chemical properties, and their contents differ from
tissue to tissue and even from location to location within a tissue. Thus, the mechanical properties
depend both on tissue and on site.
Collagen, which is defined as a protein containing sizeable domains of triple-helical confor-
mation (tropo-collagen) and is synthesized by fibroblasts, vascular smooth muscle cells, and so
on, is a basic structural element for biological tissues in animals. It gives mechanical integrity
and strength to our bodies and is present in a variety of structural forms in different tissues and
organs.
Elastin is also a long-chained protein similar to collagen, but has a much less integrated
structure than collagen. Its mechanical properties are greatly different from those of collagen and
it has much less strength and much more flexibility than collagen.
Mechanical Properties of Soft Tissues and Arterial Walls 17
0.16
Human external iliac artery
I C
I a
I
I
0.12 Trypsin/
I
I
I
I
} I
I
I
6 0.08 I
I
"'-< I
~ I
.9 I
"'~ I
~ I
I
I
I
I
0.04 I
I
I
I
I
I Formic acid
I .... b
I
/
...... ......
0 "'
1 1.2 1.4 1.6 1.8 2.0
Extension ratio 'A(=L/Lo)
Figure 2. Tension-extension ratio curves of fresh, formic acid-treated, and trypsin-treated arterial wall
(Roach and Burton (1957)).
Figure 1 shows stress-strain relations for the canine nuchal ligament, sole tendon, and intesti-
nal smooth muscle, which are rich in collagen, elastin, and smooth muscle (cell), respectively
(Hasegawa and Azuma (1974)). The elastin-rich nuchal ligament has much less strength and
much more flexibility than the collagen-rich sole tendon. The intestinal smooth muscle is much
softer than the other two tissues, and its stress-strain curve has a wide hysteresis loop, which
indicates that the tissue is viscoelastic. The elastic moduli calculated from these relations are ap-
proximately 0.4, 350, and 0.03 MPa in the nuchal ligament, sole tendon, and intestinal smooth
muscle, respectively.
2. Nonlinear Large Deformation
Biological soft tissues are geometrically and mechanically nonlinear. As can be estimated from
Figure 1, each structural component exhibits nonlinear behavior, and the nonlinearity is enhanced
by the assembly of these components into a tissue. For example, arterial wall can deform very
largely with a nonlinear relation between tension (load divided by width) and extension ratio
18 K. Hayashi
Collagen
Elastin
Smooth muscle
Figure 3. A model of biological soft tissue.
Rabbit ventral skin

0.8
o !:c. Experimental
• .& Calculated
0.6 0
Fy vs. )..y ~
0
0.4
i
1.2 1.4 1.6 1.8 2.0 1 1.2 1.4 1.6 1.8 2.0 2.2
Extension ratio )..x,).. y
Figure 4. Tension-extension ratio relations of rabbit skin (comparison of experimental data with calculated
results) (Tong and Fung (1976)).
20,----------------------------------------------,
Bovine articular cartilage
15
,-..,
~
,:l.., 10
:::B
'-'
h • ... • Experimental
"'v"' T=A[exp(BE)- 1]
.t:
r./1
0
Strain£
Figure 5. Tensile properties of articular cartilage (Woo et al. ( 1979) ).
(>.. = L / L 0 ; L =length under load; L 0 = initial length) (curve (a) in Figure 2 shows the tension-
extension curve of a fresh human external iliac artery, taken from Roach and Burton (1957)).
If the artery is digested with trypsin to selectively remove elastin from the tissue, its tension-
extension ratio curve is shifted towards left (curve (c)), indicating that the tissue becomes less
extensible. However, the slope of the curve is very similar to that of the non-treated, fresh tissue
(curve (a)) in the region of high tension. On the other hand, the curve is greatly shifted towards
the axis of extension ratio after the selective removal of collagen with formic acid (curve (b)).
The slope of curve (b) is similar to that of curve (a) in the region of small tension.
The comparison of the slopes of the three curves in Figure 2 indicates that the behavior of
arterial wall at low tension is similar to that of elastin, while the behavior at high tension is the
same as that of collagen. Such a characteristic behavior can be expressed by a mechanical model
like that shown in Figure 3.
3. Anisotropy
Since collagen and elastin are long-chained high polymers, they are intrinsically anisotropic.
Moreover, not only their fibers but also cells are oriented in tissues and organs in order that they
function most effectively. Inevitably, almost all biological tissues are mechanically anisotropic.
20 K. Hayashi
100-----------------------------------------------------.
Rabbit patellar tendon
t
•t,
(
.- 75
~
'--'
'HHH
1=:
......
0
-
~
f f f f
~
~
f
(!)
~ 50 I-
25~----~----~1------~----~1 ------~----~1 ------~--~

0 100 200 300 400
TimeT (sec)
Figure 6. Relaxation behavior of patellar tendon (Yamamoto et al. (1999)).
For example, skin has very different properties in two orthogonal directions; see Figure 4 (Lanir
and Fung (1974) and Tong and Fung (1976)); it cannot deform much in the direction of Langer's
line (Ridge and Wright (1966)), but can deform much more in the perpendicular direction. Col-
lagen fibers in the articular cartilage are preferentially oriented to the split line and, therefore,
this tissue also shows anisotropic behavior; see Figure 5 (Woo et al. (1979) ).
4. Viscoelasticity
As can be seen from Figure 1, most biological soft tissues exhibit open hysteresis loops in their
load-deformation curves, which means that they are viscoelastic. This is shown more clearly
by relaxation tests. For example, when the patellar tendon is elongated and maintained at some
length, the load does not stay at a specific level but decreases rather rapidly at first and then
gradually; see Figure 6 (Yamamoto et al. (1999)).
When we do mechanical tests on soft tissues excised from the body (in vitro experiments), we
should apply appropriate forces or deformations several times to the tissues (pre-conditioning)
so as to recover their in vivo state (Fung (1993)).
5. Strain Rate Insensitivity

Because of the viscoelastic characteristics of soft tissues, they show different properties under
different strain rates (test speed). In fact, higher strain rates give higher stresses; see Figure 7
(Vawter et al. (1978)). However, such a strain rate effect is not very large in biological soft tissues;
there are not very large differences in the stress-extension ratio curves across three orders of
magnitude of the tensile speed (CT in Figure 7). In addition, the area of the hysteresis loop (H in
Figure 7) does not depend upon strain rate. Generally speaking, therefore, biological soft tissues
8 ~----------------------------------------------------~
Canine lung parenchyma
a CT= 18 H=27
b CT=60 H=32
c CT=220 H=30
6 d CT= 900 H=28
e CT= 6500 H=35
CT = Load/unload time (sec)

H = Hysteresis (%)
1.1 1.2 1.3 1.4 1.5 1.6 1.7

Extension ratio )..
Figure 7. Tensile properties of lung parenchyma at different strain rates (Vawter et al. (1978)).
are mechanically not very sensitive to strain rate (see also the discussion on the insensitivity of
muscular arteries to strain rate contained in the chapter by HOLZAPFEL in this volume).
6. Incompressibility
Most biological soft tissues have a water content of more than 70%. Therefore, they hardly
change their volume (isovolumic) even ifload is applied, and they are almost incompressible. The
incompressibility assumption is applicable to most biological soft tissues; it has been confirmed
experimentally in arterial wall (Carew et al. (1968) and Chuong and Fung (1984)). However, it
is not the case in the articular cartilage, because the tissue is micro-porous and, therefore, water
can enter and leave pores depending upon load (Woo et al. (1979)).
The incompressibility assumption is very important in the formulation of constitutive laws
for soft tissues because the sum of all principal (logarithmic) strains is always zero.
22 K. Hayashi
1.3 Mathematical Description of Mechanical Behavior

1. Tensile Behavior
Because biological soft tissues can support large nonlinear deformations, a simple Hookean law
assuming infinitesimal strains cannot be used to describe their mechanical behavior.
Figure 8 demonstrates the tensile behavior of the human dura mater and the bovine peri-
cardium, both of which deform nonlinearly and extensively (Hayashi eta!. (1979)). If we calcu-
late the slopes of the curves at various stresses and plot them against the stresses, we obtain linear
relations between the slopes and the stresses, as shown in Figure 9. These relations are described
by
dT
d.\ =BT+C, (1)
where T is the nominal stress (engineering stress), which is obtained by dividing load by initial
cross-sectional area, A is the extension ratio and B and C are material constants.
If we integrate this equation and consider the boundary condition ofT = 0 at A = 1, we
obtain
T = A{exp[B(.\ -1)]- 1} = A[exp(BE)- 1], (2)
where E = A - 1 is a measure of strain and A = C /B.
This equation is applicable to many soft tissues, including the articular cartilage (see Fig-
ure 5). In several cases, however, we need two equations (actually two different sets of constants).
For example, the tensile behavior of aortas can be described by two equations, as can be seen
from Figure 10 (Hayashi eta!. (1981)).
2. Multi-axial Behavior
Strain-energy functions are commonly utilized for formulating constitutive laws of biological
soft tissues that undergo large (finite) deformation (Fung (1993)). Let W be the strain energy per
unit mass of a tissue, and p 0 be the density in the zero-stress state. Then p 0 W is the strain energy
per unit volume of the tissue in the zero-stress state, and this is called the strain-energy density
function. The strain energy W can be expressed in terms of the components E;j of Green strain.
In a uni-axial tensile test, the Green strain E in the tensile direction is given by
1 2
E =En = "2(.\ - 1). (3)
When such a strain-energy function exists, the components S;j of the second Piola-Kirchhoff
stress can be obtained as derivatives of p0 W (Fung (1993)). Thus,
5 . _ 8poW (4)
ZJ - OK. •
ZJ
In a simple tensile test, the only non-zero component of stress, denoted by S, is

F
S = Su = .\Ao, (5)
where F and A 0 are the load applied to a specimen and its initial cross-sectional area, respec-
tively.
2.0 , - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,
Human dura mater and I

I
bovine pericardium I
I
I
I
I
1.5 I
I
I
Human I
~ I
0-. dura mater I
::E
'-' I
I
h I
1.0
00
00
I
(!.) I
tl I Bovine
ell
I pericardium
I
I
I
I
0.5 I
I
I
/
/
/
__ ,., /
/
"'
0
1 1.05 1.10 1.15 1.20 1.25
Extension ratio A
Figure 8. Stress-extension ratio curves of dura mater and pericardium (Hayashi et a!.
(1979)).
In practical cases, the nominal stresses Tij are useful. They are given by
a (axi) ,
8poW
(6)
rij =
8aj
where ox;/Oaj are the components of the deformation gradient, and (al, a2, a3) and (xl' X2, X3)
denote the coordinates of a material particle in the zero-stress state and in the deformed state,
respectively. In a uni-axial tensile test
F
T = Tn = Ao. (7)
Equations (4) and (6) are constitutive laws or stress-strain relations. Thus, we need to know the
details of the strain-energy function. A general strain-energy function for biological soft tissues
24 K. Hayashi
60,-----------------------------------------------~
Human dura mater and
bovine pericardium
50
dT!d"A=BT+ C
-----. T=A{exp[B("A-1)] -1}
A=CjB
40
.8
30
Human
dura mater .o·
20
Bovine
pericardium
10
0 0.5 1.0 1.5 2.0

Stress T (MPa)
Figure 9. Slope of stress-extension ratio curve versus stress dura mater and pericardium (Hayashi et al.
(1979)).
is given by Fung (1973) in the form

EijEkt
W = aijkz--2-- + (f3o + f3mnpqEmnEpq) exp('yijEij + {ijktEijEkt + ... ), (8)
where O:ijkt,/3o,f3mnpq,{ij etc. are constants. This equation is not simple. However, for many
tissues it can be modified to simpler forms, since high-order terms can be neglected.
For example, skin is regarded as being in a bi-axial state, and its strain-energy function has
been given by Tong and Fung (1976) as
1 2 2 2 2 1 2 2
poW= 2(o:1E11 + o:2E22 + a3E12 + o:3E21 + 2o:4EnE22) + 2 cexp(a1E11 + a2E22
+a3Ei2 + a3E~ 1 + 2a4EnE22 + 11E~ 1 + 12E~2 + r4Ei 1 E22 + r5EnE~2 ), (9)

where a1, ... , a4, a1, ... , a4, 1'1, 1'2, 1'4, 1'5 and care constants.
From eq. (4), we obtain the stresses
apoW
Sn = oEn = a1E11 + a4E22 + cA 1 X, (10)
Different blood vessels

Descending
1.5
thoracic
aorta
Descending thoracic aorta (Upper portion)
(Lower portion)
_.._
~
p.,
~ 1.0
'-"'
,..<
"'0
~
"'0
-
<!)
0..
0
rFl
0.5
dT!d>..=BT+C
0 0.05 0.10 0.15 0.20 0.25 0.30
Stress T (MPa)
Figure 10. Slope of stress-extension ratio curve versus stress in different blood vessels (Hayashi et a!.
(1981)).
8paW
Szz = oEn = a4En + a2E22 + c A 2 X, (11)
8paW
S12 = f)E12 = a3E12 + c a3E1zX, (12)
where A 1 , A 2 , and X are functions of the strain components.

It can be seen from Figure 4 that eq. (9) is suitable for the bi-axial mechanical behavior of
rabbit skin. However, this equation is very complex. For practical purposes, the third-order terms
in the exponential function involving the constants 'Yl, ')'2 , ')'4, ')'5 are unimportant. In addition, if
one is concerned mainly with higher stresses and strains in the physiological range, and does not
care for great accuracy at very small stress level, the first group of terms in eq. (9) can be omitted,
and we have simply (Fung (1993))
26 K. Hayashi
Maxwell model Voigt model Kelvin model
F,u F,u F,u
Figure 11. Models for viscoelastic materials.
3. Viscoelastic Behavior
In Figure 11 three basic mechanical models of viscoelastic materials for uni-axial deformation
are shown, i.e. the Maxwell model, the Voigt model, and the Kelvin model. They are composed
of combinations of linear spring( s) with spring constant( s) J1, etc. and a dashpot with coefficient of
viscosity ry. A linear spring is considered to instantaneously produce a deformation u proportional
to the load F:
F=j),U. (14)
A dashpot is supposed to yield a velocity proportional to the load:
F- du (15)
- TJ dt.
Thus, the Maxwell model is formulated as

dF
du dt F
-=-+-. (16)
dt J.l TJ
The Voigt model gives
(17)
and the Kelvin model is described by

TJ dF J1,2du
F+ - - = J.L2u+ry(l + - ) - . (18)
ftl dt ftl dt
Frog sartorius muscle
100
I
I
I
I
80 I
•
ek.
,.-.... I
I
I
I
~
.9 60 I
I
"'
=
•
I
~
I
40 Relaxed,'
I
I
•
I
I
20 I
•
I
0-
.... J'
0.4 0.6 0.8 1.0 1.2 1.4 1.6
Relative length L/Lo
Figure 12. Tension-length relations of stimulated (contracted) and relaxed muscle (Aubert eta!. ( 1981 )).
1.4 Contraction of Muscle

There are two kinds of muscle in an animal body: (i) striated muscle, such as the skeletal mus-
cle and the heart muscle, and (ii) smooth muscle, as in, for example, the blood vessel. Muscle
can contract under a stimulated condition (active condition) and relax under a non-stimulated
condition (passive condition). If a muscle is stimulated while being kept at a constant length,
tension is generated (isometric contraction). On the other hand, if a muscle is stimulated under a
constant load, it shortens (isotonic contraction). Both behaviors are called contraction. Examples
of the relations between tension and length of a muscle under stimulated and relaxed conditions
are demonstrated in Figure 12 (Aubert et al. (1981 )). Tension generated by contraction increases
with muscle length, reaches a maximum, and then decreases; under relaxed condition, tension in
a muscle increases monotonically with length like other connective tissues.
2 Mechanical Properties of Arterial Walls

Arterial wall mechanics is indispensable not only for the understanding of blood circulatory
physiology but also for analyzing the mechanisms of vascular diseases. This section deals with
28 K. Hayashi
Location I TA I AA I CA I FA I AL I CL I VL I vc I PA I
Diameter ( mm) 13 9 5 4 0.05 0.006 0.04 10 17
Thickness ( mm) 0.65 0.5 0.3 0.4 0.02 0.001 0.002 0.15 0.2
Total area (mm 2 ) 200 200 300 300 12500 60000 57000 300 230
Peak velocity (mm/sec) 1050 550 - 1000 7.5 0.7 3.5 250 700
Reynolds number 3400 1250 - 1000 0.09 0.001 0.035 700 3000
TA = Thoracic aorta, AA = Abdominal aorta, CA = Carotid artery, FA = Femoral artery, AL =

Arteriole, CL =Capillary, VL =Venule, VC = Vena cava, PA =Pulmonary artery
Table 1. Arterial size and blood flow parameters in the canine (reproduced from Caro et al. (1978)).
the basic elastic properties of arterial wall and their changes with age, including the mathematical
formulation of the mechanical properties.
Arterial wall is composed of three layers (intima, media and adventitia), which are separated
by elastic membranes (Hayashi et al. (2001)). The luminal surface is lined with a layer of en-
dothelial cells, which are in contact with blood. Shear stress developed by blood flow is applied
to the endothelial cells. The media supports load induced by blood pressure, and its mechanical
properties represent the properties of arterial wall. Further discussion of arterial histology and the
mechanical characteristics of arterial components is contained in the chapters by HUMPHREY
and MCCULLOCH and by HOLZAPFEL in this volume.
I Location A TA AA I CIA I FA CCA PA

R(mm) 10.8 7.4 4.5 2.6 3.1 4.4 14.3
±i1R/R (%) 2.7 1.5 1.0 0.2 0.6 0.5 5.4
Ep (10 5 N/m2 ) 0.74 1.20 2.00 9.82 4.33 6.08 0.16
R = Radius, Ep = Pressure strain elastic modulus, A = Ascending aorta, TA

=Thoracic aorta, AA =Abdominal aorta, CIA= Common iliac artery, FA=
Femoral artery, CCA =Common cartoid artery, PA =Pulmonary artery
Table 2. Pulsatile change of wall radius inside the body, ±i1R/ R, and wall stiffness, Ep (Patel et al.
(1964)).
Arterial wall is mainly composed of elastin, collagen, and cells (endothelial cells, smooth
muscle cells, and fibroblasts); the composition varies between locations depending upon there-
quired function (Hayashi et al. (2001)). For example, collagen (C) and smooth muscle (SM)
increase and elastin (E) decreases at more distal locations in conduit arteries; see Figure 13
(Hayashi et al. (1973)). Inevitably, the ratio of collagen to elastin (C/E) increases in more distally
located arteries. Collagen and elastin are essentially similar proteins, but collagen is very much
stronger and stiffer than elastin (see Figure 1). Therefore, the diameter changes of arteries devel-
'T
2.0
Canine artery
cam •
+--E 1.6
@l
~
,.....::,~
UVJ
.._,
s::
.._,
ll)
ll)-
•··· j:)
~
~
u
.._,
:a"' s"'
btl C)
/
30 1.2 .g
u.s::
<...,+->
0 0
I
.-"
------
.-" /
~
.....
.st;
...... 0
s ...., ,,,,, +--c .-".-" /
-
s::
E "' .......... .,.,.....-" /
/
~"'
'i:l
................,.
S::"O
0 s:: 20 _,d
0.8 ll)
C) "' btl
ll) •
btl~ u... .!
C/E~
"'~
...... .._,
l 0
t
s:: s::
IU·~ ............ \:!""."' u
u+->
..... "' 10-
1
ll) " '
p..~
0
CCA TA AA CIA
Figure 13. Compositions of arterial wall (CCA = Common carotid artery, TA = Thoracic aorta, AA = Ab-
dominal aorta, CIA= Common iliac artery) (Hayashi et al. (1973)).
oped by blood pressure pulsation depend upon their sites; the diameter change is larger in more
proximal arteries, see Table 2 (Patel et al. (1964)). Smooth muscle cells can contract and relax,
which serves to control blood pressure.
2.1 Arterial Structure and Composition

The blood circulatory system is divided into the systemic circulation, pulmonary circulation, and
coronary circulation. In the systemic circulation, arterial blood pumped out from the left ven-
tricle reaches capillaries via aortas, arteries, and arterioles, and then returns to the right atrium
via venules, veins, and vena cavas after being changed to venous blood in capillaries. The ve-
nous blood is then pumped out from the right ventricle into the pulmonary artery and finally
returns to the left atrium via pulmonary veins after oxygenation in alveoli in the lung (pulmonary
circulation). In the coronary system, arterial blood enters coronary arteries from the aortic root
immediately after departing the left ventricle, and perfuses the heart wall.
The diameter and wall thickness of blood vessels are different at locations in the vascular
tree depending on functional demands, see Table 1 (Caro et al. (1978)). In general, the arterial
diameter and thickness gradually decrease at more distal locations, but the total cross-sectional
area increases with the distance from the heart and becomes maximal in the capillaries.
2.2 Pressure-Diameter Relation

A variety of methods have been used to determine the mechanical properties of blood vessels,
including tensile tests of dumbbell-shaped specimens and ring specimens, and pressure-diameter
30 K. Hayashi
Experimental system for arterial pressure-diameter relations
Recorder
Video
Pressure dimension
transducer analyzer
Diaphragm Krebs-Ringer _ } Specimen

solution (37°C}
Figure 14. An experimental system for arterial pressure-diameter relations (Hayashi (2000)).
tests of tubular specimens (Hayashi et at. (2001)). The pressure-diameter tests are preferable to
the tensile tests, because the tubular shape of blood vessels inside the body is preserved.
An example of the experimental devices which are used for tubular specimens is shown in
Figure 14 (Hayashi (2000)). In this system, pressure inside the specimen is measured with a
pressure transducer, specimen diameter with a video dimension analyzer via a CCD camera, and
axial load with a strain gage-mounted cantilever-type load cell. The internal pressure or specimen
diameter can be controlled with a sequencer during the inflation, deflation and contraction of a
blood vessel.
The left panel of Figure 15 demonstrates typical pressure-radius relations obtained under
normal conditions (solid curve) and under an active condition (dashed curve) (Hayashi et at.
(1980b )). These data were obtained while the specimen was maintained at the in vivo length
(in vivo axial extension ratio, see below) that was measured inside the body before excision.
The relation under the normal condition was obtained in a Krebs-Ringer solution which is anal-
ogous to the normal environment inside the body (physiological condition). The dashed curve
was obtained in the Krebs-Ringer solution administrated with 10- 5 molar norepinephrine which
contracts arteries (active condition or contracted condition).
Under the normal conditions, arteries have nonlinear pressure-radius (or diameter) relations;
they greatly increase the diameter with pressure under a low pressure range, say below 60 mmHg,
and then gradually lose the distensibility at higher pressures. When arteries are contracted un-
der an active condition, their diameter decreases in a physiological pressure range and below
the range (below 200 mmHg in Figure 15), and their pressure-radius (or diameter) curves be-
2.5 , - - - - - - - - - - - - - - - ,
200 Human femoral artery/
/ 2.0
• 00
•
/
Diameter response /
.··o
'eo =!::.Ro/Ro(N) /
/
.o
/ 0
1150 / c;i
Active // Normal (2{ Active
0... (contracted)// (~=11.2) o.P.. (~=3.2) o
?~
~
"' 100
"'~
-0..
ro
E I
:s
<!)
50 I
I
I
0
0
•In(P/ Ps)=~(R 0 / Rs- I)
• ~=Stiffuess parameter
0 CL__ _~--~--~--~--~~~
2.5 3.0 3.5 Ro(N) 4.0

R~(N)
0.5
0.8
0
0.9
•
"---~---'--~----"--~_L__~___L~----"
1.0 1.1 1.2 1.3
External radius R 0 (mm) Distension ratio ( R0 / Rs)
Figure 15. Pressure/radius (left) and pressure ratio/distension ratio (right) relations of a human femoral
artery determined under normal and active (contracted) conditions (Hayashi eta!. (1980b)).
come very different from those observed under the normal condition. The contractile response
is often expressed by the diameter response iJ.D 0 / D0 (N) (or similarly by the radius response
iJ.R0 / Ro (N)), where iJ.D 0 and iJ.R0 are the changes in diameter and radius, respectively, at
each pressure, while D 0 (N) and Ro(N) are the diameter and radius measured at the pressure
under the normal condition, respectively (see Figure 15).
Inside the body, almost all aortas and arteries are tethered to the surrounding tissues. There-
fore, when we excise arteries from the body, they shorten. The ratio of arterial length inside the
body to that measured after excision (in vivo axial extension ratio) is roughly 1.5, which de-
pends upon arterial site, age, disease, and so on. Thus, when we perform in vitro mechanical
experiments, we have to account for this.
2.3 Mechanics of Arterial Wall
1. Arterial Stiffness
For practical purposes, when we measure the elastic properties of arterial wall, it is convenient
to use a single parameter that expresses arterial elasticity under living conditions rather than to
use such perfect, but complicated formulations as the constitutive equations mentioned below
in detail. For this purpose, several parameters have been proposed to express arterial elasticity,
which include the 'pressure-strain elastic modulus Ep' (Peterson et al. (1960)) and the 'vascular
32 K. Hayashi
compliance Cv' (Gow and Taylor (1968)). These parameters are, respectively, described by
(19)
and
LlV
c- LlP'
v-
v (20)
where Do and V are the outer diameter and the luminal volume, respectively, of a blood vessel
at pressure P, and LlD0 and Ll V are their increments for a pressure increment LlP.
These parameters are defined at specific pressures and give different values at different pres-
sure levels, because the pressure-diameter relations of arteries are nonlinear, as mentioned above.
To overcome this inconvenience, several functions have been proposed to describe such relations
mathematically. Among these functions, the following equation is one of the simplest and most
reliable for the description of pressure-diameter (or radius) relations of arteries in the physiolog-
ical pressure range (Hayashi eta!. (1980a)):
P
ln-=(3
Ps
(Do
- - 1)
Ds
or P (Ro
ln-=(3
Ps
- - 1).
Rs
(21)
Here, Ps is a standard pressure, and D 8 and Rs are the vascular diameter and radius, respectively,
at the pressure Ps. A physiologically normal blood pressure such as 100 mmHg is recommended
for P 8 • The right panel of Figure 15 shows the close fit of the data to eq. (21) over a rather
wide pressure range under normal conditions. This equation is also applicable to the pressure-
diameter relations under active conditions within some limited pressure range, as can be seen
from Figure 15.
The coefficient (3 in eq. (21 ), called the stiffness parameter, represents the structural stiffness
of a vascular wall. This parameter has been used for the evaluation of the stiffness of arteries not
only in basic investigations, but also in clinical studies (Hayashi (1993)). As stated above, one
of the most important merits of this parameter is that it does not depend on the pressure in the
physiological pressure range. At the standard pressure P 8 , the parameters described by eqs. (19)
and (20) are converted into the stiffness parameter as follows:
(22)
2. Elastic Modulus
To represent the elastic properties of wall material, it is necessary to use a material parameter
such as the so-called elastic modulus or Young's modulus, which is the slope of a linear stress-
strain relation. For arterial walls that have nonlinear stress-strain relations, incremental elastic
moduli have been used for this purpose. For example, the incremental elastic modulus
(23)
where Di is the internal diameter of a vessel, has been proposed (Berge! (1961 )). This modulus
was derived from the classical Lame equation for the infinitesimal inflation of a long, thick-
walled tube made of an idealized linear elastic, isotropic, and incompressible material. Actually,
these assumptions are not always valid for blood vessels.
Later, Hudetz (1979) modified the formulation and proposed a more reasonable incremental
elastic modulus for an orthotropic cylindrical vessel having a nonlinear stress-strain relation. This
is described by
2 i1P 1 2 1
Hee = 2Di Do i1D D 2 _ D~ + 2PD 0 D 2 _ D 2 . (24)
0 0 1 0 I
To calculate these moduli, we need to know the thickness or internal diameter of a vessel. In
in vitro experiments, we can calculate them from D 8 , internal and external diameters under no
load condition, and the in vivo axial extension ratio, assuming incompressibility of the wall ma-
terial. Noninvasive measurements of wall thickness or internal diameter on intact vessels inside
the body are not easy, although it is now possible with high-accuracy ultrasonic echo systems
(Hayashi et al. (2001)).
3. Constitutive Laws
Under the in vivo condition, arteries are subjected to both internal pressure and axial force. In in
vitro experiments, these forces and arterial dimensions (length, internal and external diameters,
and wall thickness) can be determined precisely. These data are useful for the mathematical
modelling of the mechanical properties of arterial wall which aims to determine constitutive
equations that relate quantities describing stress and strain states. For an elastic solid material,
the constitutive equations have the form of stress-strain relations, which can be derived from a
potential function called the strain-energy function W (Fung (1993) and Hayashi (1993)). Thus,
(25)
where Eij are the Green strains. Since the derivatives of W with respect to the strains give stress-
strain relations, as described in the previous section (eqs. (4) and ( 6)), the problem of determining
constitutive equations is reduced to the determination of strain-energy functions.
Three major equations have so far been proposed for the strain-energy function of arterial
wall. In Vaishnav et al. ( 1973) the equation
was advocated, where Po is the density in the zero-stress state, Eee and Ezz are Green strains in
the circumferential and axial directions, respectively, and A, ... , G are constants.
Fung et al. ( 1979) proposed a simpler form with an exponential function in the form
(27)
where d, a1, a2, and a4 are constants.

Later, Takamizawa and Hayashi ( 1987) proposed a logarithmic form of the function described
by
(28)
34 K. Hayashi
50
Canine femoral artery
- - - l- - - - - - - i
T
40
Mean pressure
----------------------_I 140mmHg
-
en_
...
11) 30 '~
I
11)
§
til
0.
tl.l
lOOmmHg
tl.l
1
11)
@ 20
·.;::
t ---!- ------t ----------- I

V1
60 mmHg
10
Pulse pressure=40 mmHg
'-------~·--· L_ ~__c___-~~- ·~~1__~~~~ ~____L___~~ _j_--~-- ___]_____
0 1 2 3 4
Frequency f(sec~ 1)
Figure 16. Effects of the frequency of pulsatile pressure on arterial stiffness (Kotera and Hayashi (1981)).
where C, aoo, a 22 , and aoz characterize the mechanical properties of a material. From this equa-
tion, stress-strain relations are obtained as
_ _ C (1 + 2Eoo)(aooEoo + aozEzz) (29)

ao ar- (1 -tJr) ,
(30)
where
(31)
and a o, a z, and a r are Cauchy stresses in the circumferential, axial, and radial directions, respec-
tively. Additionally, constitutive laws for the elastic and viscoelastic behavior of arterial walls,
which were proposed recently, are discussed in the chapter by HOLZAPFEL in this volume.
Although all of the proposed formulations describe quite well the elastic behavior of arterial
walls, we prefer to reduce the number of constants included in the equations in order to handle
them more easily, as well as to give physical meanings to the constants. For this reason, the
exponential (eq. (27)) or logarithmic (eq. (28)) expressions are advantageous.
Canine saphenous artery
NE=5·10- 6 M
.'''
KCI=50mM
. ''
.....
.~
250
.l
•'
' Normal
r/J
.,
: :
d :: c :'b a
·:
... ...
'
1:'..
I
.. ..
'
'''
. ' /! j:'
.s'..
'' '
.... ..
'
' ' '
.. ' '
.. '
.'. ..
'
'
'' ..'
I I
' ' I ~ I
"
:.· .·
1 1 I I I
;' ,'
I I
::
I
II I I
I I
I I
I
:
II I I II
,1: : ..
11 I I I
II I I
50 II I II
II I If
II I tl
.. I ,f
l :
,' I I I
I I II
l I I I I
1/ S, I"'
0 0.8 1.0 1.2 1.4 1.6 1.8

External diameter D0 (mm)
Figure 17. Effects of the degree of contraction on arterial pressure-diameter relations (NE =Norepinephrine)
(Nagasawa et al. (1980)).
2.4 Effects of Pulsation and Contraction
Arterial stiffnesses measured under pulsed pressure conditions are different from those obtained
under quasi-static pressurization, see Figure 16 (Kotera and Hayashi (1981 )). Under pulsed pres-
sure conditions, however, stiffness does not depend upon the frequency of pressure pulsation,
which may reflect the insensitivity of the mechanical properties of biological soft tissues to strain
rate, as mentioned in the previous section. These phenomena have also been observed in human
arteries (Learoyd and Taylor (1966)).
If arteries are stimulated physically or chemically, they contract and change their pressure-
diameter relations (Figure 15). The degree of contraction greatly affects the shape of the rela-
tions; see Figure 17 (Nagasawa et al. (1980)). If arteries that are rich in smooth muscle cells
are moderately contracted, their pressure-diameter curves show open hysteresis loops. If they are
contracted a lot, the loops are closed, and the curves become almost straight.
2.5 Locational Differences and Aging Changes of Arterial Stiffness
Figure 18 shows ,8-values of common carotid arteries, intracranial vertebral arteries, and coro-
nary arteries obtained at autopsy from human subjects of different ages (Gow and Hadfield ( 1979)
36 K. Hayashi
60~--------------------------------------------------~
Human artery
• _. • Hayashi et al. o Gow and Hadfield
0
40
en_
....
1-<
<l)
<l) Coronary artery
~
til 30
0.
"'"'
<l)
§....
VJ
20
Intracranial vertebral artery
10
Common carotid artery
0 20 40 60 80
Age Y(Year)
Figure 18. Age-related changes in arterial stiffness (in vitro studies) (Hayashi et a!. (1986) and Gow and
Hadfield (1979)).
and Hayashi et al. (1986)). It should be noted that arterial stiffness is much larger in the coronary
arteries than in the other arteries, and also that intracranial vertebral arteries are considerably
stiffer than extracranially located common carotid arteries. Such locational differences in (arte-
rial) wall stiffness Ep are also seen from Table 2.
Almost all the data obtained from normal human aortas and arteries show that the structural
stiffness of wall (for example, Ep and /3) increases or the wall compliance Cv decreases with
age. As can be seen from Figure 18 (Hayashi et al. (1986) and Gow and Hadfield (1979)) and
Figure 19 (Kawasaki et al. (1987) and Reneman et al. (1986)), wall stiffness in such conduit
arteries as the common carotid artery and aortas increases with age rather gradually until the age
of 40 years, and very rapidly thereafter (Hayashi (1993)). The stiffness of intracranial arteries
such as the intracranial vertebral artery progressively increases until 20 years, and then more
slowly (Hayashi et al. (1980a)). There seems to be almost no change in the human coronary
artery with age.
50,-----------------------------------------------------,
Human aorta
•
en_
40 • Thoracic aorta
(Reneman et al.) ••••
•
.-
1-<
......
<!)
<!)
til 0 Abdominal aorta

•
t<J
0.. 30- (Kawasaki et al.)
en
en
•
<!)
,:
~
.....
......
• •
CZl
20-
·"'J•c • •
10 f-
••
~~~
•. ~
0
8fS:DfP fa•~· •ooo o o

I I _l I
0 20 40 60 80 100
Age Y(Year)
Figure 19. Age-related changes in aortic stiffness (in vivo measurements) (reproduced from Kawasaki eta!.
(1987) and Reneman eta!. (1986)).
3 Mechanical Properties of Cells
It is now well recognized that cells change their shape, structure, mechanical properties, and
function in response to mechanical stress. Because cytoskeletal structure is closely related to cell
function and is directly reflected in the mechanical properties of cells, the determination of cell
properties should contribute much to the study of the mechanisms of tissue and organ physiology,
diseases, and other events that occur in the body. Moreover, it is basically very important to know
the mechanical properties of cells in order to understand cell physiology.
Various methods and techniques have been applied to the determination of the mechanical
properties of cells. These are divided into two categories: (i) measurement of the local properties
in a cell, and (ii) measurement of the properties of a whole cell. Local mechanical properties
have been studied on a variety of cells using the techniques of, for example, cell poking (Zahalak
et al. (1990)), micropipette aspiration of local cell surface layer (Sato et al. (1987a)), twisting
of embedded (Valberg and Feldman (1987)) or surface-attached (Wang and Ingber (1994)) mag-
netic particles with external magnetic fields, bending of an extended portion of an adherent cell
with micro-needles (Albrecht-Buehler (1987)), scanning acoustic microscopy (Bereiter-Hahn et
38 K. Hayashi
Micropipette
Figure 20. Endothelial cell aspirated into micropipette (.:1P = Suction pressure, R = Pipette internal radius,
L =Aspirated cell length) (Sato et al. (1987a)).
al. (1995)), and atomic force microscopy (Hoh and Schoenenberger (1994) and Miyazaki and
Hayashi (1999)). The mechanical properties of a whole cell have been studied mainly on blood
cells and muscle cells using the methods of micropipette aspiration of a whole cell (Hochmuth
et al. (1993)), compression and stretch of a whole cell with a pair of microplates (Thoumine and
Ott ( 1997)), and stretch of a whole cell with a pair of micropipettes (Glerum et al. ( 1990), Palmer
et al. (1996) and Miyazaki et al. (2000)).
From these studies, this section deals with three techniques: (i) a micropipette technique and
its application to cultured endothelial cells, (ii) an atomic force microscopic method for intact
endothelial cells in fresh arteries, and (iii) a tensile test of a whole cell.
3.1 A Micropipette Technique and Its Application to Cultured Endothelial Cells
A micropipette having the internal diameter of approximately 2.7 J.Lm is filled with a culture
medium (MDM), connected to a pressure control line, and then fixed to the stage of a microscope
(Sato et al. (1987a) and Sato et al. (1987b )). The pressure control system is composed of two
different reservoirs, the heights of which are adjustable. A difference between these two heights
can produce a negative pressure at the inside of the tip of the micropipette (Figure 20). One of
these reservoirs is open to the atmosphere and can be controlled with an accuracy of better than
10 mm by the use of a micrometer. The other reservoir is connected with the micropipette using a
fine tube. A part of the tube between the micropipette and the reservoir is connected to a pressure
transducer for the measurement of the pressure level in the micropipette system.
Endothelial cell (EC) suspension is loaded into a cell chamber (2 mm height, 10 mm width)
which is held by a manipulator. The suspended cells are observed through a long working dis-
tance objective lens under the microscope. The tip of the micropipette is made to approach the
surface of a spherical EC by manipulating both the cell chamber and the stage of the microscope.
When a negative pressure is applied to the tip of the micropipette, the cell is drawn toward and
slightly into the pipette (Figure 20). First, a negative pressure of about 2 mm H 2 0 is set by using
the micrometer, and then a portion of the EC surface is aspirated into the micropipette. In pre-
liminary experiments, the aspirated part of the surface continued to deform after the application
of pressure, but an almost steady state was observed to occur within 8 to 10 minutes. Therefore,
the negative pressure is maintained for 10 minutes at each setting, and the image of the aspirated
portion of the cell is taken with a TV camera connected to the microscope. The length of the
aspirated portion of the surface in the micropipette L is measured at several different negative
pressures !J.P during the loading process. After the aspirated portion of the cell surface has at-
tained a length of twice the radius of the micropipette, the negative pressure is decreased and
the cell is unloaded. The recovery of the deformed portion of the cell surface is measured during
the unloading process. When the pressure is returned to zero, it is checked whether the aspirated
portion has fully recovered.
The data in the form of L versus !J.P represents, in effect, a stress-strain relationship. The
data are presented here in the form of a mechanical stiffness parameter K, defined by
K = R!J.P (32)
L '
R
where R is the micropipette radius, R!J.P is the tension being exerted, and L / R is the strain
(non-dimensional aspirated cell length) imposed on the cell.
Fully confluent cultured bovine aortic ECs of the 7th to 9th generation were used to study
the effect of fluid-imposed shear stress on the stiffness of the cells (Sa to et a!. (1987b )). The
cells were exposed to a steady shear stress (laminar fluid flow) using a parallel plate flow cham-
ber. Briefly, cultured ECs were positioned in the central part of the flow chamber which has a
flow section of 220 mm in height, 17 mm in width, and 50 mm in length. The chamber, one
of the reservoirs, and the circulation circuit were filled with culture medium (MDM contain-
ing 25 mmol/1 Hepes buffer, 20% fetal bovine serum, and antibiotics) at 37°C. The ECs in the
flow chamber were exposed to a specific shear stress condition defined by the dimensions of the
chamber and the pressure drop across the chamber.
Figure 21 demonstrates typical loading data of the ECs exposed to shear stresses of 1, 3,
and 8 Pa for 24 hours as well as of a control cell subjected to no flow for the same duration.
The relations between pressure difference and aspirated length are approximately linear. The
slope of such a relationship is used as the stiffness parameter K (see eq. (32)). After 24 hours
of exposure to shear stresses, the increase in the stiffness parameter was statistically significant
as compared with control values for all the three shear stress levels employed. In addition, the
stiffness parameter was larger at higher shear stresses.
In this study, Sato et al. (1987b) also observed that the end face shape of the ECs on the
substrate became more elongated and their long axis became oriented to the direction of flow.
40 K. Hayashi
Vascular endothelial cells exposed to

various levels of shear stress
Shear stress (Pa)
• 8
,-...,
s
~
c..
<l
~ 2 3
Q)
(.)
::::
Q)
1-<
~ 1
:0
Q)
1-<
::l
ell
ell
~
p.. 1
0 0.5 1.0 1.5 2.0 2.5

Aspirated length L / R
Figure 21. Effects of shear stress on force (stress)/deformation (strain) relation of vascular endothelial cells
cultured under laminar flow for 24 hours (Sato et al. (1987b)).
There was also an alteration in the F -actin filaments. These results suggest that the influence of
shear stress on EC mechanical stiffness is related to alterations in the cytoskeletal structure.
It might be expected that at an early stage of atherosclerosis the cytoskeleton of ECs could
play an important role in determining the permeability of endothelium to macromolecules, for
example, via transendothelial and junctional transport. Cells in high shear regions, in response to
their mechanical stress environment, would have a more highly developed cytoskeletal structure
which could inhibit both transendothelial and junctional macromolecule transport. Thus, the re-
sults shown in Figure 21 imply that low shear regions, with their higher influx of macromolecules
such as low density lipoproteins (LDLs), would have a higher predilection for the initiation of
atherosclerotic lesions. The response of ECs to shear stress includes an alteration in both its
structure and mechanical properties. More detailed comparisons of structure and cell function
between in vivo and cultured ECs are needed to apply the results of the above-mentioned in vitro
observations to the ECs in living blood vessels.
Abdominal aorta
6mm
Lateral wall
Common iliac artery
Medial wall
Figure 22. Locations of the arterial segments used for atomic force microscopic measurements of the stiff-
ness of endothelial cells (Miyazaki and Hayashi (1999)).
3.2 Atomic Force Microscopic Method for Intact Endothelial Cells in Fresh Arteries
Atomic force microscopy is a useful technique for imaging biological specimens such as cells,
proteins, and DNA with high resolution without any specific treatment of samples (La! and John
(1994)). The measurement oflocal mechanical properties of a cell is also possible with a nanoin-
dentation technique using atomic force microscopy (Weisenborn et al. (1993)). The mechanical
properties of several kinds of cultured cells, for example, MDCK cells (Hoh and Schoenenberger
(1994)), myocytes (Shroff et a!. (1995)), carcinoma cells (Goldmann and Ezzell (1996)), 3T6
cells (Ricci eta!. (1997)), have been studied using this method. To our knowledge, however,
there has been no report on the atomic force microscopic measurement of the mechanical prop-
erties ofliving cells in situ. Thus, Miyazaki and Hayashi (1999) have determined the mechanical
properties of living endothelial cells (ECs) in fresh vascular segments obtained from rabbit aortic
bifurcations (Figure 22) using an atomic force microscope (AFM), and studied their locational
differences. After the aortic bifurcation in the rabbit was carefully exposed, the external diam-
eters of the abdominal aorta and both sides of the common iliac arteries were measured with
calipers. Then each vessel was marked with gentian violet at 5 mm intervals on the outer surface
along the axial direction to identify the in vivo axial length. After sacrifice, the aortic bifurcation
between the position proximal from the posterior mesenteric artery and that distal to the internal
iliac arteries was excised, and immediately immersed in Hanks' balanced salt solution (HBSS)
at room temperature and then stored at 4°C.
Prior to AFM observation, strip specimens having the axial and circumferential length of
approximately 5 mm and 3 mm, respectively, were cut from the lateral and medial sites of the
42 K. Hayashi
Atomic force microscopy
Cantilever holder
\
Silicone tube
37°C
HBSS (+)
--~
\ --~
Sample stage
Coverslip
(Sample)
Cantilever
X-Y-Z piezo scanner
Figure 23. Stiffness measurement by means of atomic force microscopy.
bifurcated common iliac arteries and from the anterior wall of the abdominal aortas (Figure 22).
Each specimen was placed on a silicone rubber (15 x 15 x 3 mm) with the endothelial side
up, and the endothelium was covered with a droplet of HBSS to prevent it from drying. Then,
the specimen was fixed to the rubber with tiny stainless steel pins, being stretched to its in vivo
axial and circumferential length. The thicknesses of the specimens thus attached were 110 pm
to 130 pm. Immediately after being soaked in HBSS at room temperature for a few minutes,
the specimen attached to the silicone rubber was mounted onto the sample stage of an AFM
(Figure 23).
The AFM equipped with a large-area piezo scanner having a maximum x-y scanning range
of 100 x 100 pm and a z range of 15 pm was used for the study. V-shaped silicon nitride can-
tilevers having a very tiny pyramidal tip at the end were used not only for imaging endothelial
surfaces but also for measuring the mechanical properties of ECs. First, the surface topography
of the endothelium was observed in a contact mode at an x-y scanning range of 50 x 50 pm
and a frequency of 1 Hz. A droplet of HBSS was often put on the endothelial surface during the
observation. The imaging force was kept as low as possible (0.04 to 0.06 nN) to avoid damage of
ECs. After the surface topography, force-indentation relations were obtained by applying force
to individual ECs by moving the sample stage upward at a rate of 440 nmjsec. Force was auto-
matically calculated from the cantilever deflection measured with a laser beam and a photodiode,
while indentation was obtained from the difference between the movement of the sample stage
and that of the cantilever tip caused by cantilever deflection.
AFM images of the endothelium demonstrated that ECs in the abdominal aorta and in the
medial wall of the aortic bifurcation were elongated and oriented to the axial direction of each
vessel, while they were round in the lateral wall of the bifurcation. The topographic profile of the
2.0 Force-indentation curves

in an endothelial cell
1.5
,--.._
~
"-'
l::t;
<1.)
1.0
(.)
.....
0
~
0.5
0 0.2 0.4 0.6 0.8 1.0

Indentation 5 (f.Lm)
Figure 24. Force-indentation curves obtained from different locations in an endothelial cell on a rabbit ab-
dominal aorta (Miyazaki and Hayashi (1999)).
endothelium in the anterior wall of the abdominal aorta was similar to those in the medial and
lateral wall of the aortic bifurcation; the peak-to-bottom height of ECs ranged between 0.5 and
1.5 Jllli.
A wide variety of force-indentation curves were obtained from different locations in a cell
(see Figure 24), which indicates that endothelial cells are structurally and mechanically inho-
mogeneous. The slopes of force-indentation curves obtained from the locations over cell nuclei
seemed to be greater than those from more peripheral locations. The highest point of each EC
was determined from the topographic profile using the software incorporated in the AFM sys-
tem, and a force-indentation curve was obtained from the point. The scattering of the curves thus
obtained from different cells was much less than that observed at different positions in a cell
(Figure 25). Based on these results, force-indentation data were obtained from the highest points
of cells for comparison among the vascular sites.
Force-indentation data obtained from different cells at each location of individual bifurcations
(see, for example, Figure 25) were first averaged, and then the data from all the bifurcations
(animals) were averaged for each location. Thus, averaged force-indentation curves obtained at
the three locations (see Figure 22) are shown in Figure 26. All three curves are nonlinear, and
the curve from the lateral wall is very similar to that from the aorta. The slope of each curve
increases with an increase in applied force; it is larger in the medial wall of the iliac artery than
44 K. Hayashi
2.0 Force-indentation curves

of endothelial cells
1.5
,.-._
~
'-'
k.
~
()
1-<
1.0
0
~
0.5
0 0.2 0.4 0.6 0.8 1.0
Indentation 5(1-1m)
Figure 25. Force-indentation curves obtained at the highest point in different endothelial cells on a rabbit
abdominal aorta (Miyazaki and Hayashi (1999)).
in the abdominal aorta and the lateral wall of the iliac artery when compared at the same force.
There were significant differences in the indentation amongst the locations at and below 0.25 nN.
The mechanical properties of endothelial cells were evaluated from the force-indentation (F-
8) relation
F =a [exp(M)- 1], (33)
where a and bare constants. The slope of the force-indentation curve is
dF
M = bF + ab = bF + c. (34)
The parameters b and c (= ab) correspond to the rate of modulus change and initial modulus,
respectively, and are used to express the stiffness of cells. In particular, the parameter b is related
to structural inhomogeneity and represents its change induced by stress. The parameter a is used
as an index to evaluate the shape of force-indentation curves; this parameter is related to the
locational change of inhomogeneous structure inside a cell.
The parameter a calculated from force-indentation relations was significantly higher in the
medial wall than in the lateral wall; on the other hand, no significant difference was observed be-
tween the aorta and the lateral wall (Figure 27). There were almost no differences in the parameter
0.6~------------------------------------------~
Average force-indentation curves of

endothelial cells at three locations
0.5
0.4
,-...,
~
'-'
k;
0
(.)
1-<
0.3
•. Abdominal
aorta
*
0
~
Lateral
wall
0.2
* • Medial
wall
*
0.1 Mean±S.D. (n = 6)
*p < 0.05 (One-way ANOVA)
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Indentation 6 (1-1m)
Figure 26. Average force-indentation curves obtained at the highest point in endothelial cells at three loca-
tions in rabbit aortic bifurcations (Miyazaki and Hayashi (1999)).
b between the three locations. There were significant differences in the parameter c among the
locations by one-way ANOVA; it was higher in the medial wall than in the other two locations.
These results indicate that ECs in the medial wall of the iliac artery are stiffer than those in the
other sites, which may reflect the existence of abundant stress fibers in the ECs in the medial
wall. The highly developed cytoskeletal structure with abundant stress fibers in the medial wall,
which is probably caused by high wall shear, serves as a barrier against the uptake of low density
lipoproteins (LDLs) in the wall, and prevents the formation of atherosclerotic plaques (Sato et al.
(1987b )). On the other hand, the relatively undeveloped cytoskeletal structure ofECs in low shear
regions promotes LDL intake, which would be a reason for the predilection for atherosclerosis
in the lateral wall.
3.3 Tensile Testing of a Whole Cell
Simple tensile testing of a whole cell is useful for determining the basic mechanical properties.
However, there have been only a few studies on the tensile properties of cells. Thoumine and Ott
( 1997) and Palmer et al. ( 1996) studied the viscoelastic and contractile properties of fibroblasts
46 K. Hayashi
Parameter values calculated from force-indentation relations of endothelial cells
Mean±S.D.
0.4 p < 0.02 (t-tes~ 1.4 p < 0.05
(One-way ANOVA)
I 8 1.2
~
I
0.3
~ 1.0
"' ~ 0.8
0.2 ~
~ 0.6
Po. 0.1 ~ 0.4
Po. 0.2 i
0 L_i__ _ l _ L _ L L L _ L _ L - L 0 c___L_ ____l_LlLLLLLL
Aorta Lateral Medial Aorta Lateral Medial Aorta Lateral Medial
(n= 6) (n= 6) (n = 6) (n= 6) (n = 6) ( n = 6) (n = 6) (n = 6) (n = 6)
Figure 27. Parameter values calculated from force-indentation relations at three locations in the rabbit aortic
bifurcation (Miyazaki and Hayashi (1999)).
and cardiac myocytes, respectively, and also described their tensile properties. However, they
did not determine tensile force-elongation relations or the tensile strength of these cells. Glerum
et a!. (1990) determined the tensile properties of smooth muscle cells obtained from the pig
urinary bladder and human uterus by knotting the ends of a single cell around the tips of a pair
of micropipettes and moving one of the micropipettes. Their knotting technique, however, can be
applied only to relatively long and stiff cells. Thus, Miyazaki eta!. (2000) have recently designed
a tensile test system for cells and applied it to the determination of the tensile properties of
fibroblasts. The system has also been used for smooth muscle cells (Hayashi (2000) and Miyazaki
eta!. (2001)).
The tensile test system is composed of a thermostatic test chamber, an inverted fluorescence
microscope, micromanipulators, a direct drive linear actuator, a cantilever-type load cell, and a
video dimension analyzer (VDA) (Figure 28). The test chamber made of acrylic plates is de-
signed with a window of glass coverslip at the bottom. It is mounted on the inverted fluorescence
microscope. The temperature of a solution in the chamber can be kept at 37°C with the exter-
nal circulation of heated water and with a heater placed on the stage of the microscope. The
microscope is placed on a vibration isolator.
The load cell consists of a very thin stainless-steel cantilever (0.03 mm thick) and a laser
displacement meter; the cantilever is housed in a small case (20 x 20 x 10 mm) and attached
to an acrylic arm which is attached to a piezoelectric micromanipulator. The load applied to
a specimen is obtained from the deflection of the cantilever, which is measured with the laser
displacement meter. The calibration of the load cell is performed using a reference glass micro-
needle having a known spring constant.
A pair of micropipettes (20 to 30 J.tm outer diameter, 3 to 5 J.tm inner diameter) used for
gripping a cell were fabricated from glass capillary tubes (1 mm outer diameter, 0.56 mm inner
diameter) using a micropipette puller. After bending each micropipette at two positions, their
fine tips used for cell gripping were coated with a cell adhesive (Cell-Tak). One of the double-L-
shaped micropipettes is attached to the load cell, and its thick end is connected to a microsyringe
via a silicone tube (1 mm outer diameter, 0.5 mm inner diameter) and a glass capillary tube
(Figure 29). The micro syringe is connected to a micro-injector. The other micropipette is attached
Tensile test system for cells
Microscope
Cantilever CCD
Micromanipulator Thermostatic chamber
Monitor Micromanipulator
Video timer Actuator
Video recorder
VDA
LDM Vibration isolator

Tube pump
CLSM Thermostatic bath
PC = Personal computer
VDA = Video dimension analyzer
LDM = Laser displacement meter
CLSM = Confocal laser-scanning microscope
Figure 28. Tensile test system for cells - overall view (Miyazaki et a!. (2000, 2001) and Hayashi et a!.
(2000)).
to a hydraulic micromanipulator which can be moved by the linear actuator, and its thick end is
connected to another microsyringe via a Teflon tube.
The pair of micropipettes were filled with Hanks' balanced salt solution (HBSS), and were
attached to the tensile test system. HBSS (37°C, pH 7.4) containing floated cells were poured
into the test chamber. The image of a cell was recorded on videotape to measure its diameter
under non-loaded condition using an image analyzer. The cell was then attached to the fine tip
of one of the micropipettes which was connected to the load cell, by applying very low negative
pressure produced by the microsyringe using the micro-injector (Figure 29). Care was taken not
to suck the surface layer of the cell into the micropipette. The pressure application was continued
for 15 to 20 minutes to firmly bond the cell to the micropipette tip making use of the cell adhesive
coated on the tip. The silicone tube between the micropipette and the micro syringe was cut gently
to apply the testing load directly to the load cell. Then, the fine tip of the other micropipette was
moved to the cell using the micromanipulator and attached to the opposite side of the cell in the
same manner as that mentioned above.
After returning internal pressure in the micropipette to atmospheric pressure, the cell was
stretched until fracture at a rate of 6 J.Lm/sec using the linear actuator. During the tensile test-
ing, the images of the cell and the micropipette tips were monitored through the differential
interference optical system and recorded on a video cassette recorder. Elongation of the cell is
48 K. Hayashi
Details of a tensile test system for cells
Windshield casing
I
Silicone tube Condenser
Micropipettes
Outer diameter = 20 - 30~-tm
Inner diameter = 3 - 5~-tm
Coated with cell adhesive (Cell-Tak)
Figure 29. Details of a tensile test system and a cell attached to micropipettes (Miyazaki et a!. (2000, 200 I)
and Hayashi eta!. (2000)).
obtained from the distance between the fine tips of the two micropipettes using the VDA via
the microscope and a CCD camera. The maximum stroke, displacement rate, accuracy of force
measurement, and resolution of displacement measurement of this system are 10 mm, 1Jlm/sec
to 10 mm/sec, ±0.05 J1N, and 0.24 Jlm, respectively.
Collagen fascicles (300 Jlm diameter) were resected from the patellar tendon of matured
Japanese white rabbits. Fibroblasts (FBs, 20.6 ± 3. 7Jlm diameter in mean±S.D.; n = 6) were
isolated from these fascicles by enzymatic digestion using collagenase. Smooth muscle cells
(SMCs, 32.3 ± 7.4 Jlm diameter in mean± S.D.; n = 6) were obtained from the thoracic aorta
of matured Japanese white rabbits with an explant method. These SMCs are categorized into a
synthetic phenotype.
FBs and SMCs have a similar shape ofload-elongation curve; see Figure 30 (Miyazaki et al.
(2000) and Miyazaki et al. (2001)). However, SMCs are much stiffer and have higher strength
than FBs. We could not determine the breaking load of SMCs because they detached from the
micropipette tips before cell fracture occurred. Both cells have the same function of synthesizing
protein molecules for connective tissues. However, their structures are very different, which may
have reflected the great difference in their tensile properties.
4 Biomechanics of Vascular Diseases
The mechanical properties of arterial walls influence not only the blood circulation but also the
development and progression of arterial diseases via effects on blood flow and arterial mass trans-
port. Vascular diseases change the wall properties. Furthermore, stress and strain in the arterial
wall are extremely important in the understanding of the pathophysiology in the vascular system.
Elastic modulus 17 29% 47% 24%

Stiffness 17 71% 29% 0%
Wall thickness 12 67% 25% 8%
Table 3. Distributions of reported data of the elastic modulus, stiffness and wall thickness of atheroscle-
rotic wall (reproduced from Hayashi (1993)).
From the topics of biomechanics in vascular diseases, this section deals with: (i) atherosclero-
sis and arterial elasticity, (ii) wall elasticity of hypertensive arteries, and (iii) cerebral arterial
diseases.
4.1 Atherosclerosis and Arterial Elasticity
The effects of flow dynamics and wall shear stress on the initiation and development of atheroscle-
rosis have been studied extensively (Nerem (1992)). However, less attention has been paid to the
mechanical properties of atherosclerotic wall tissue. Does atherosclerosis stiffen the arterial wall
or increase the elastic modulus of wall tissue? The results obtained have been conflicting and
inconclusive, as shown in Table 3 (Hayashi (1993)). One of the reasons for this is that the struc-
tural stiffness of arterial wall and the elasticity of wall material have been used, confusingly, for
the expression of the elastic properties of atherosclerotic wall. In Table 3, the elastic modulus
represents the elasticity of wall material, which corresponds to the slope of stress-strain curve,
and is given by, for example, the incremental elastic modulus Einc or Hoe (see eqs. (23) or (24));
the stiffness is the structural stiffness expressed by, for example, the stiffness parameter j3 (see
eq. (21)), the pressure-strain elastic modulus Ep (see eq. (19)), or the inverse of the vascular
compliance Cv (see eq. (20)), (Hayashi (1993) and Hayashi et al. (2001)).
Several studies have shown that the arterial wall is stiffened by the development of atheroscle-
rosis. For example, Richter and Mittermayer (1984) reported that the vascular compliance Cv of
autopsied human aortas was lower at the more advanced stage of atherosclerosis, although the
compliance at around 100 mmHg increased at an early stage of the disease having minimal le-
sions (Figure 31 ). However, several studies have presented different results. Hudetz et al. (1981)
demonstrated that there were no differences in the wall stiffness between fibrosclerotic human
cerebral arteries and normal ones, although the incremental elastic modulus was lower in fi-
brosclerotic walls (Figure 32). Thus, it is not clear whether atherosclerosis increases the elastic
modulus of arterial wall. From Table 3 we can see that atherosclerosis is mostly accompanied by
wall thickening. This might be a reason why there are no data indicating a decrease in the struc-
tural stiffness associated with atherosclerosis. The structural stiffness is determined not only by
the elastic modulus of wall material but also by wall dimensions such as wall thickness (Hayashi
(1993) and Hayashi et al. (2001)).
The lipids incorporated into arterial walls and the calcification of a wall are considered to
soften and stiffen the atherosclerotic wall, respectively. This combination might have yielded
such inconclusive results. To solve this problem, we need precise data for the mechanical prop-
erties of atherosclerotic lesions and plaques. However, there have been only a few studies on
the stiffness and strength of atherosclerotic plaques, and the results reported are variable and
50 K. Hayashi
3.0,---------------------------------------------~
Fibroblasts and vascular

smooth muscle cells
2.5
Mean±S.D. (n = 6)
2.0
Smooth muscle cell

(D0 =32.3±7.41JID)
1.5
1.0 Breaking point
0.5
Fibroblast (D0=20.6±3.7~J.m)
0 20 40 60 80 100 120
Elongation L (~J.m)
Figure 30. Load-elongation relations of fibroblast and vascular smooth muscle cells (Miyazaki et a!. (2000,
2001)).
20 , - - - - - - - - -
Human aorta
~ 15 Degree of atherosclerosis
~
~
0 Normal
I Minimal
2 Moderately severe
rS 3 Severe
<!)
(.)
10
-~
]'
0
(.)
1(.)
~"'
........
0 50 100 150 200 250 300

Internal pressure P (mmHg)
Figure 31. Vascular compliance in different degrees of atherosclerosis (Richter and Mittermayer (1984)).
15 , - - - - - - - - - - - - - 10 •
Human internal
carotid artery
12 - 8-
o Normal
0
• Fibrosclerotic •
9- CJ)
0
0 •
••
6 r-
0
0 ••
c9 •
0 oo • •'
0 •••
OJ •
,~···I I _j_
0 50 100 150 200 250 0 50 100 150 200 250
Internal pressure P (mmHg)

Figure 32. Incremental elastic modulus and vascalar compliance of normal and fibrosclerotic human internal
carotid arteries (Hudetz et al. (1981 )).
inconclusive, possibly due to different methods of testing, different species, variations in the mi-
croscopic structure of plaques, anisotropy, etc. (Humphrey (1995) and Hayashi et al. (2001)).
For a more precise and detailed study of the mechanical properties of atherosclerotic plaques,
Hayashi and Imai (1997) used a statistically meaningful number of specimens (10 mm length,
1.5 mm width, less than 0.5 mm thickness) obtained from the rabbit thoracic aorta. Atheroscle-
rosis was induced by the denudation of endothelial cells in the aorta with intraluminal ballooning
and the feeding of cholesterol diet. The stress-strain curves of the plaques and wall media showed
nonlinear relations and large deformation (Figure 33). The plaques have much smaller slopes than
the wall media, which was essentially similar to the results obtained by Castle and Gow (1983).
The tensile strength of the plaques was 131 kPa on average, and this strength was much lower
than that of wall media. Therefore, atherosclerotic walls could be softer and weaker than healthy
walls.
To make clear the relation between atherosclerosis and wall elasticity, Hayashi et al. ( 1994)
have performed a detailed and systematic study of the mechanical properties and morphology of
atherosclerotic aortas in the rabbit. The results indicated that the changes in the wall stiffness (3
(see eq. (21)) and the elastic modulus Hoe (see eq. (24)) are not always correlated with the time of
cholesterol diet feeding for up to 32 weeks. Thus, the grade of atherosclerosis was defined from
the percent fraction of the luminal surface area stained with Sudan IV as well as from wall stiffen-
ing. The area fraction of intimal hyperplasia Ah increased with the grade (Figure 34). Likewise,
wall thickness steadily increased with the progression of atherosclerosis (Figure 35). However,
the elastic modulus was not significantly different from the control artery until the highest grade
of atherosclerosis. On one hand, there appeared a significant increase in the arterial stiffness in
the grade II atherosclerosis, which is attributable to the wall thickening. Significantly increased
52 K. Hayashi
Media and plaque of rabbit

thoracic aortas
200
,....,
o::l
~
'-'
h 150
X Breaking point
"'"'
~
C/l
100
Extension ratio >..

Figure 33. Tensile properties of wall media and atherosclerotic plaque of rabbit thoracic aortas (Hayashi and
Imai (1997)).
calcification and intimal hyperplasia were observed in the wall of the grade III atherosclerosis.
From these results, it is concluded that the progression of atherosclerosis induces wall thicken-
ing, followed by wall stiffening. However, even if atherosclerosis is advanced, there is essentially
no change in the elastic modulus of wall material unless considerable calcification occurs in the
wall. Calcified aortas have high elastic moduli. At the most advanced stage of atherosclerosis, the
arterial wall has high structural and material stiffness due to calcification and wall hypertrophy.
Stress developed in the arterial wall by blood pressure is a very important factor for arterial
diseases. For example, once atherosclerotic plaques and atheroma are formed, they should have
a strong influence on the distribution of stress and strain in the wall. If wall stress is high, it may
affect the progression of atherosclerosis. High stresses in plaques and atheroma may result in
their disruption.
Hayashi and Imai ( 1997) calculated stress distributions in atherosclerotic walls by means of
a finite element method. Considering the morphology of atherosclerotic plaques and atheroma,
they used four different models: (i) two models for local plaque and (ii) two for uniform atheroma.
The inner and outer diameters of parent vessels (media) were assumed to be the same for all the
models. Models A and B were designed to represent local plaques formed on the intraluminal
surface of vessels in the early stage of atherosclerosis. The maximum thickness of the plaques
in Models A and B were designed to be 1/3 and 2/3 of the wall thickness of the parent ves-
sel, respectively. Models C and D were axisymmetric two-layer cylindrical models consisting
70
Artherosclerotic thoracic aortas
of rabbits
60
Denudation of endothelial cells
and cholesterol diet
50
til
s,.....,
..... -;f.
"d'-'..c::
..... 40
.... < (Mean±SE)
~0 .....
"'
.....0 "'"'
~e. 30
c.!:: !.)
"'~
~.c:
20
10
0 L__,----~-
0 II III
Grade of atherosclerosis
Figure 34. Area fraction in% of intimal hyperplasia in atherosclerotic thoracic aortas of rabbits (reproduced
from Hayashi eta!. (1994)).
of atheroma and wall media, in which the wall luminal surface was uniformly lined with the
atheroma. This type of atherosclerosis is observed at the interim or early end stage of the dis-
ease. The stress-extension ratio relations shown in Figure 33 were used for wall media, and for
plaques and atheroma. For the calculation, a strain-energy density function proposed by Klosner
and Segal (1969) was applied to these relations; the agreements between experimentally obtained
stress-extension ratio data and calculated results were very good.
In the aortic wall attached by a local plaque (Models A and B), circumferential tensile stress
was maximal in the wall media very near the plaque edge, and the stress value was very high
compared with the tensile strength of wall media (Table 4). This local stress concentration may
be a contributing factor to the progression of atherosclerotic disease because such high stresses
could stimulate cells in the wall. In the aortic wall uniformly lined with atheroma (Models C and
D), circumferential stress was high in the innermost layer of the atheroma, and its magnitude
was comparable to the tensile strength of plaque. The stress gradient at the interface between the
atheroma and the wall was also high. These results indicate a possibility of rupture of atheroma
near the intraluminal surface of atheroma or at the junction between atheroma and wall intima.
For a layer-specific finite element model of an individual human stenosis, which considers three-
dimensional morphological data coming from high-resolution magnetic resonance imaging and
54 K. Hayashi
* p < 0.02 (vs. control)

Mean±SE
Stiffness
(ratio vs. control)
1t___.-~
6~---------------------------------------
Elastic modulus
(ratio vs. control)
4
2
1
Area fraction of
calcified region (%)
0 control 0 I II III
Grade of atherosclerosis
Figure 35. Elasticity, wall thickness, and calcification of athorosclerotic thoracic aortas in rabbits (repro-
duced from Hayashi eta!. (1994)).
PI queor Wall
Atheroma media
Mu MiD Mu MID
Local plaque models
Model A: Along plaque center line 72 0 292 142
Near plaque edge - - 677 -
Model B: Along plaque center line 80 0 264 132
Near plaque edge - - 870 -
Uniform atheroma models
Modele 125' 87'. 244' 97 ..
Model D 141' so·· 153' 6s··
Tensile strength 131±39 approx. 1.000
Units in (kPa), * at the inner layer, ** at the outer layer
Table 4. Calculated stresses in atherosclerotic aorta models (Hayashi and lmai (1997)).
associated histological analyses, the reader is referred to the chapter by HOLZAPFEL in this
volume.
4.2 Wall Elasticity of Hypertensive Arteries
Hypertension is recognized as an important risk factor for many cardiovascular diseases, includ-
ing atherosclerosis and cerebral hemorrhage. Elevated blood pressure exerts influences on the
synthetic activity of vascular smooth muscle cells, and is believed to induce changes in struc-
ture and morphology of the arterial wall, its mechanical properties, and vascular contractility. It
is therefore very important to understand arterial mechanics in hypertension. However, results
from the extensive literature concerning the mechanical properties of hypertensive arteries are
contradictory and inconclusive (Hayashi (1993), Humphrey (1995) and Hayashi et al. (2001)).
Several studies were performed to determine the pressure-diameter or pressure-volume rela-
tionships of aortas and arteries in hypertensive animals and humans (Hayashi et al. (200 1)). For
example, Greenwald and Berry (1978) have shown that, at a given pressure, the aortas from the
rats exposed to induced hypertension for 6 and 20 weeks were structurally stiffer than those from
normotensive animals. On the other hand, Vaishnav et al. (1990) reported that the aortas from
dogs in which hypertension had been induced for 2 and 4 weeks were slightly more distensible
than normotensive aortas at comparable intravascular pressures. More recently, Matsumoto and
Hayashi (1994) performed a more detailed study on the effects of induced hypertension on the
pressure-diameter relationship and elastic modulus of the rat thoracic aorta. Comparison of hy-
pertensive animals with normotensive controls showed that at 100 mmHg and also at the working
pressure (systolic blood pressure Psys before sacrifice) of each group, the pressure-strain elastic
modulus Ep (see eq. (19)), was greater in hypertensives than in normals, whereas at 200 mmHg
56 K. Hayashi
Structural stiffness of thoracic aortas
-
80
D Ep at 100 mmHg f'sys=Systolic blood pressure
rJl
.E 70
l2Zj Ep at f'sys * p<0.05 vs. normotensive
.§ Ep at 200 mmHg (Mean±SE)
0 60
8
()
·~~ 50
,g::c:
.s ~
ell ' - '
40
.!:J 0..
~kl 30
~
rJl
rJl 20
....
<!)
~
10
0
2wks 4wks Swks 16wks 2wks 4wks Swks 16wks
Normotensive Hypertensive
(Psys = 134 mmHg) ( Psys = 197 mmHg)
Figure 36. Structural stiffness of thoracic aortas in normotensive and hypertensive rats (reproduced from
Matsumoto and Hayashi (1994)).
the Ep values in the hypertensive animals were slightly lower than those of the normals (Fig-
ure 36). These results did not depend upon the duration of hypertension for 2 to 16 weeks. When
we analyze the reported data, we should remember that the values of such parameters as Ep and
Cv (see eq. (20)) are dependent upon pressure. Without this consideration, comparisons between
the results from different studies have little meaning.
With regard to the inherent elastic modulus of wall material calculated from pressure-diameter
data, Greenwald and Berry (1978) reported that, at physiological levels of pressure and above,
the incremental elastic modulus Einc (see eq. (23)) of the aorta was lower in hypertensive rats
than in normotensive ones when compared with a given pressure or circumferential stretch ratio.
Even if compared with in vivo systolic blood pressure levels, the incremental elastic modulus
was lower in hypertensive rats than in normotensive animals. They ascribed the result to the rel-
atively higher content of elastin and lower content of collagen in hypertensive animals than in
normotensive ones. It should be noted that in this study hypertension was induced in animals
aged only 4 weeks and that the response of the young vessel to increased pressure may differ
from that of its mature counterpart. On the other hand, Vaishnav et al. (1990) observed that the
stress-extension ratio relationship of the thoracic aorta determined from pressure-diameter-axial
force-length data was very similar in the circumferential direction in hypertensive and normoten-
sive dogs; however, the hypertensive aorta had a higher slope than its normotensive counterpart
in the longitudinal direction.
Elastic modulus versus blood pressure in the rat aorta

10,---------------------------, 10,--------------------------,
2wks • 8 8 wks
~ 6
.a
rJl
4
..§ 2 2
~~ OL_l_~_L_L~_ _L_l_~_L_L~~ OL_l_~_L~~--L_~J__L_i~~
.~~-
~. 100 125 150 175 200 225 250 100 125 150 175 200 225 250
10,-----------------------------,
•
O:S CD
~ ~ 4wks 16wks
:
8
~ 8
6 r=0.59
~
(N.S.)
g
4
• ••
...... 2
OL_L_L_~i_~_L_L_L_L~~~
2
OL_L_L_~~~_L_L_L~~~~
• •
100 125 150 175 200 225 250 100 125 150 175 200 225 250
Systolic blood pressure Psys (mmHg)
Figure 37. Incremental elastic modulus versus systolic blood pressure in the rat aorta at 2, 4, 8 and 16 weeks
after the treatment for hypertension (Matsumoto and Hayashi (1994)).
Matsumoto and Hayashi (1994) have shown that the incremental elastic moduli Hoe (see
eq. (24)) at systolic blood pressure levels Psys had significant correlations with blood pressure
until 8 weeks after the induction of hypertension; at 16 weeks, however, the correlation disap-
peared and the elastic modulus tended to be at the same level as that in control, normotensive
rats (Figure 37). Based on these results, they concluded that the aortic wall in hypertensive rats
restored the in vivo elastic properties to a normal level in 16 weeks due to the functional adapta-
tion and remodelling of the wall. They also observed no significant differences in the incremental
elastic modulus at 100 mmHg regardless of the period of hypertension.
In connection with the mechanical properties of wall, many data have shown that the arterial
wall was thickened by hypertension, although a few results have suggested no changes (Hayashi
et al. (2001)). Wall thickness depends critically upon pressure level and, therefore, we have to
pay attention to the pressure for the thickness measurement when we interpret the results. Only
Matsumoto and Hayashi ( 1994) determined the wall thickness at the in vivo blood pressure level.
They used the wall thickness to calculate in vivo wall stress in the circumferential direction,
and observed that the stress was independent of the degree of hypertension and was always
maintained at a control, normal level even at 2 weeks after the induction of hypertension. They
ascribed this phenomenon to a functional adaptation and remodelling of the arterial wall.
4.3 Diseases of Cerebral Arteries
Mechanical properties of cerebral arteries are very important for the study of the pathogenesis
of cerebrovascular lesions such as aneurysms, vasospasm, and atherosclerosis. For example, the
formation and rupture of saccular aneurysms, which are common in cerebral arteries and rather
58 K. Hayashi
40,--------------------------------------------,
Wall stiffness of human arteries
-
CJ
30 Young (below 45 yrs)
en_
..... Old (above 45 yrs)
E<l)
~
0.. 20 (Mean±SE)
"'"'
<l)
~
".;::l
rfJ
10
BA VA VA ICA CCA
Intracranial Extracranial
Figure 38. Wall stiffness of human intracranial and extracranial arteries in two age groups (BA = Basilar
artery, VA= Vertebral artery, ICA =Internal carotid artery, CCA =Common carotid artery (re-
produced from Hayashi eta!. (1980c)).
rare elsewhere, are considered to be a breakdown of the arterial wall caused by the force induced
by blood pressure and blood flow (see also the chapter by HUMPHREY in this volume, which
focuses on intracranial saccular aneurysms). Cerebral vasospasm, which occurs very frequently
after the ictus of subarachnoid hemorrhage and then results in a serious ischemic condition in
the brain, is a phenomenon of the reduction of arterial diameter (arterial contraction). Full under-
standing of the mechanical properties of cerebral arteries is needed to study the mechanisms and
mechanics of these diseases.
Human intracranial arteries, which are stiff already at birth, start increasing their wall stiff-
ness soon after birth, and become significantly stiffer than extracranial arteries at the age of
20 years (Hayashi et al. (1980a)). Figure 38 demonstrates the stiffness (3 (see eq. (21 )) of five
kinds of human arterial walls, two of which were obtained from intracranial sites (Hayashi et al.
(1980c)). The subjects were divided into two age groups: (i) 'Young' for less than 45 years of
age and (ii) 'Old' for over 45 years. This age grouping was determined considering that systemic
arteries start stiffening at 45 years (see Figures 18 and 19, and Hayashi (1993)). Figure 38 shows
higher wall stiffness in intracranial arteries than in extracranial arteries in both age groups. High
stiffness in the intracranial arteries of the 'Young' group corresponds to the common occurrence
of cerebral aneurysms not only in the elderly but also in the middle-aged persons. A model study
on the effect of arterial stiffness on pulsatile blood flow using polymer tubes with different dis-
300,-------------------------------------------------~
Pressure diameter relations of intracranial vertebral arteries
Subarachnoid Control (Tumor)

250 hemorrhage (50 yrs)
(50 yrs)
,.-._
OJ)
200
~
'--'
c.,
0
!3 150
"'"'
~
A
-s
c;j
0
100
......
~
0
2.0 2.5 3.0 3.5 4.0 4.5 5.0
Figure 39. Pressure-diameter relations under active (KCl) and passive (saline) conditions of intracranial ver-
tebral arteries excised from human subjects with subarachnoid hemorrhage and brain tumor (con-
trol) (Nagasawa eta!. (1980)).
tensibility indicated that the stiffer the tube was, the higher was pulsatile pressure (Moritake et
al. (1974)). Such large fluctuations of pressure may occur in stiffer cerebral arteries compared
with extracranial arteries, which may induce degenerative changes and structural weakness in the
wall of cerebral arteries, especially at the apices of arterial bifurcations. Such high stiffness of
intracranial arteries may play an important role in the pathogenesis of cerebral aneurysms, and
their growth and rupture.
Cerebral vasospasm often occurs after subarachnoid hemorrhage. Biomechanical analyses of
the phenomenon are also very important. Examples of pressure-diameter curves of the intracra-
nial vertebral arteries obtained from human autopsied subjects with subarachnoid hemorrhage
(SAH) and with brain tumor (Control) are shown in Figure 39 (Nagasawa et al. (1980)). The
curves obtained in saline solution are very similar in shape to each other, showing little differ-
ence in the wall stiffness (3 between the two arteries. The activation of vascular smooth muscle
with KCl caused vasoconstriction and shifted the curves toward the pressure axis. The curve of
the control artery is biphasic and has a flexion at the pressure of 20 mmHg; the wall is stiff be-
low the flexion point and becomes more distensible at higher pressure. The curve of the artery
subjected to SAH has a flexion point at a pressure of as high as 180 mmHg, below which the
wall shows very little distension. It is suggested from these data that blood elevation clinically
60 K. Hayashi
300,---------------------------,
Control SAH model
Serotonin D
D
:: •
cf I
I
•
r-
D
I •
D
I I
D
: ••
a
r-B
D
Krebs-
Ringer + I •
a I ••
D
D ,.•••,.. •• •
• ••Saline
Saline
~
0
0.9 1.1 1.3
_j
1.5 0.9
I
·~ -
T
1.1
J
1.3
I I
1.5
Figure 40. Pressure-diameter relations under active (serotonin), normal (Krebs-Ringer) and passive (saline)
conditions of non-treated, control basilar artery (left) and the artery of an experimental subarach-
noid hemorrhage model (SAH model, right) in the canine (Nagasawa eta!. (1982)).
induced to retain blood flow for relieving cerebral ischemia caused by vasospasm is effective
only if blood pressure is maintained above the pressure level of the flexion point.
To determine the mechanisms of cerebral vasospasm, an experimental study was carried
our on the basilar artery harvested from dogs which were intracisternally injected with a small
amount of blood in vivo (SAH model) (Nagasawa et al. (1982) and Nagasawa et al. (1983)). The
pressure-diameter data of the artery obtained from the animals that underwent blood injection
into the cistern (SAH model in Figure 40) are similar to those observed in the human subject
with SAH (Figure 39). The stiffness parameter (3 and the incremental elastic modulus Einc of
the canine basilar artery significantly decreased after the SAH treatment, having their minimum
values at 2 days (Figure 41 ). These parameters gradually increased thereafter, but did not recover
to the control levels even at 4 weeks. The temporal changes in the arterial stiffness and wall
elasticity were in parallel to the change in the ratio of the content of collagen to that of elastin.
Therefore, the reduction of wall stiffness observed after the treatment results from the de-
crease of the content ratio of collagen to elastin, because collagen is much stiffer than elastin.
The reduced wall stiffness possibly makes smooth muscle, which is a counterpart component
of collagen and elastin, contract more easily. That is, the lower the stiffness of the arterial wall,
the more efficient the wall constriction produced by the activation of smooth muscle. The reduc-
tion of stiffness and the decrease of the relative content of collagen in the arterial wall subjected
to SAH might be one of the factors affecting the development of vasospasm. In addition, rela-
tively high contractility of cerebral arteries (Nagasawa et al. (1979)) may also promote cerebral
vasospasm.
Stiffness parameter and elastic modulus

of the canine basilar artery
20,----------------------, 2.0,-----------------
,-.,
::s""
15 A. 1.5
C!l.
.... '-'
.....
~
~
u
.:
~
~
10 "' 1.0
.E
0..
"'"' .g
0
~
§..... s
.~
CZl .....
"' 0.5
G:l"" P=lOOmmHg
(Mean±SE) (Mean±SE)
0 0
c 5 10 15 20 25 30 c 5 10 15 20 25 30
Period after blood injection T (Day)

( c=control)
Figure 41. Stiffness parameter and incremental elastic modulus of the canine basilar artery in an experimen-
tal subarachnoid hemorrhage model (Nagasawa et al. (1982)).
References
Albrecht-Buehler, G. (1987). Role of cortical tension in fibroblast shape and movement. Cell Motif.
Cytoskel. 7:54-67.
Aubert, X., Roquet, M. L., and van der Elst, J. (1981). The tension-length diagram of the frog's sartorius
muscle. Arch. Int. Physiol. 59:239-241.
Bereiter-Hahn, J., Karl, I., Luers, H., and Voth, M. (1995). Mechanical basis of cell shape: Investigations
with the scanning acoustic microscope. Biochem. Cell Bioi. 73:337-348.
Berge!, D. H. (1961). The static elastic properties of the arterial wall. J. Physiol. 156:445-457.
Carew, T. E., Vaishnav, R. N., and Patel, D. J. (1968). Compressibility of the arterial wall. Circ. Res.
23:61-68.
Caro, C. G., Pedley, T. J., Schroter, R. C., and Seed, W. A. (1978). Mechanics of the circulation. Oxford
Univ. Press.
Castle, W. D., and Gow, B. S. (1983). Changes in the microindentation properties of aortic intimal surface
during cholesterol feeding of rabbits. Atherosclerosis 47:251-261.
Chuong, C. J., and Fung, Y. C. (1984). Compressibility and constitutive equation of arterial wall in radial
compression experiments. J. Biomech. 17:35-40.
Fung, Y. C., Fronek, K., and Patitucci, P. (1979). Pseudoelasticity of arteries and the choice of its mathe-
matical expression. Am. J. Physiol. 237:H620-H631.
Fung, Y. C. (1973). Biorheology of soft tissues. Biorheology 10:139-155.
Fung, Y. C. (1993). Biomechanics. Mechanical Properties of Living Tissues. New York: Springer-Verlag,
2nd edition.
62 K. Hayashi
Glerum, J. J., Mastrigt, R. V., and Koeveringe, A. J. V. (1990). Mechanical properties of mammalian single
smooth muscle cells. III. Passive properties of pig detrusor and human a terme uterus cells. J Muscle
Res. Cell Motif. 11:453--462.
Goldmann, W. H., and Ezzell, R. M. (1996). Viscoelasticity in wild-type and vinculin-deficient (5.51)
mouse F9 embryonic carcinoma cells examined by atomic force microscopy and rheology. Exp. Cell
Res. 226:C234-237.
Gow, B. S., and Hadfield, C. D. (1979). The elasticity of canine and human coronary arteries with reference
to postmortem changes. Circ. Res. 45:588-594.
Gow, B. S., and Taylor, M.G. (1968). Measurement of viscoelastic properties of arteries in the living dog.
Circ. Res. 23:111-122.
Greenwald, S. E., and Berry, C. L. (1978). Static mechanical properties and chemical composition of
the aorta of spontaneously hypertensive rats: A comparison with the effects of induced hypertension.
Cardiovasc. Res. 12:364-372.
Hasegawa, M., and Azuma, T. (1974). Wall structure and static viscoelasticities of large veins. J Jap.
College Angiol. 14:87-92 (in Japanese).
Hayashi, K., and Imai, Y. (1997). Tensile property of atherosclerotic plaque and an analysis of stress in
atherosclerotic wall. J Biomech. 30:573-579.
Hayashi, K., Sato, M., Handa, H., and Moritake, K. (1973). Biomechanical study of vascular walls (test-
ing apparatus of mechanical behavior of vascular walls and measurement of volume fraction of their
structural components). Proc. 16th Jap. Cong. Mat. Res. 240-244.
Hayashi, K., Kiraly, R. J., and Nose, Y. (1979). Mechanical evaluation of storage treatment of natu-
ral tissues as valve materials. Artif. Organs 3, Suppl. (Proc. 2nd Meet. Int. Soc. Artif. Organs, New
York):417--422.
Hayashi, K., Handa, H., Nagasawa, S., Okumura, A., and Moritake, K. (1980a). Stiffness and elastic
behavior of human intracranial and extracranial arteries. J Biomech. 13: 175-184.
Hayashi, K., Nagasawa, S., Naruo, Y., Moritake, K., Okumura, A., and Handa, H. (1980b). Parametric
description of mechanical behavior of arterial walls. J Jap. Soc. Biorheology 3:75-78.
Hayashi, K., Nagasawa, S., Naruo, Y., Okumura, A., Moritake, K., and Handa, H. (1980c). Mechanical
properties of human cerebral arteries. Biorheology 17:211-218.
Hayashi, K., Washizu, T., Tsushima, N., Kiraly, R. J., and Nose, Y. (1981). Mechanical properties of aortas
and pulmonary arteries of calves implanted with cardiac prostheses. J Biomech. 14: 173-182.
Hayashi, K., Igarashi, Y., and Takamizawa, K. (1986). Mechanical properties and hemodynamics in coro-
nary arteries. In New Approaches in Cardiac Mechanics, K. Kitamura, H. Abe and K. Sagawa (Eds).
Tokyo: Gordon and Breach. 285-294.
Hayashi, K., Ide, K., and Matsumoto, T. (1994). Aortic walls in atherosclerotic rabbits- Mechanical study.
ASME J Biomech. Eng. 116:284-293.
Hayashi, K., Stergiopulos, N., Meister, J.-J., Greenwald, S. E., and Rachev, A. (2001). Techniques in the
determination of the mechanical properties and constitutive laws of arterial walls. In Leondes, C., ed.,
Cardiovascular Techniques Vol. 1!, Biomechanical Systems Techniques and Applications, 6--61. Boca
Raton: CRC Press.
Hayashi, K. (1993). Experimental approaches on measuring the mechanical properties and constitutive
laws of arterial walls. ASME J Biomech. Eng. 115:481--488.
Hayashi, K. (2000). Biomechanics. Tokyo: Corona (in Japanese).
Hochmuth, R. M., Ting-Beall, H. P., Beaty, B. B., Needham, D., and Tran-Son-Tay, R. (1993). Viscosity
of passive human neutrophils undergoing small deformations. Biophys. J 64: 1596--1601.
Hoh, J. H., and Schoenenberger, C. A. (1994). Surface morphology and mechanical properties ofMDCK
monolayers by atomic force microscopy. J Cell Sci. 107:1105-1114.
Hudetz, A. G., Mark, G., Kovach, A. G. B., Kerenyi, T., Fody, L., and Monos, E. (1981). Biomechanical
properties of normal and fibrosclerotic human cerebral arteries. Atherosclerosis 39:353-365.
Hudetz, A. G. (1979). Incremental elastic modulus for orthotropic incompressible arteries. J Biomech.
12:651~655.
Humphrey, J.D. (1995). Mechanics of the arterial wall: Review and directions. Crit. Rev. Biomed. Eng.
23:1~162.
Kawasaki, T., Sasayama, S., Yagi, S., Asakawa, T., and Hirai, T. (1987). Non-invasive assessment of the
age related changes in stiffness of major branches of the human arteries. Cardiovasc. Res. 21:678-687.
Klosner, J. M., and Segal, A. (1969). Mechanical characterization of a natural rubber. PIBAL Report.
69-42, Polytechnic Inst.
Kotera, H., and Hayashi, K. (1981). A study on the dynamic mechanical behavior of arterial wall. J Jap.
Soc. Biorheology 88~91 (in Japanese).
La!, R., and John, S. A. (1994). Biological applications of atomic force microscopy. Am. J Physiol.
266:Cl-C21.
Lanir, Y., and Fung, Y. C. (1974). Two-dimensional mechanical properties of rabbit skin: II. Experimental
results. J Biomech. 7: 171 ~ 182.
Learoyd, B. M., and Taylor, M. G. (1966). Alterations with age in the viscoelastic properties of human
aortic walls. Circ. Res. 18:278~292.
Matsumoto, T., and Hayashi, K. (1994). Mechanical and dimensional adaptation of rat aorta to hyperten-
sion. ASME J Biomech. Eng. 116:278~283.
Miyazaki, H., and Hayashi, K. (1999). Atomic force microscopic measurement of the mechanical proper-
ties of intact endothelial cells in fresh arteries. Med. & Bioi. Eng. & Comput. 37:530~536.
Miyazaki, H., Hasegawa, Y., and Hayashi, K. (2000). A newly designed tensile tester for cells and its
application to fibroblasts. J Biomech. 33:97~104.
Miyazaki, H., Hasegawa, Y., and Hayashi, K. (2001). Tensile properties of vascular smooth muscle cells.
Proc. 2001 Bioeng. Conf -ASMEBED-Vol. 50:155~156.
Moritake, K., Handa, H., Okumura, A., Hayashi, K., and Niimi, H. (1974). Stiffness of cerebral arteries-
Its role in the pathogenesis of cerebral aneurysms. Neurologia Medico-Chirurgica 14-1 :47~53.
Nagasawa, S., Handa, H., Okumura, A., Naruo, Y., Moritake, K., and Hayashi, K. (1979). Mechanical
properties of human cerebral arteries: Part 1 Effects of age and vascular smooth muscle activation.
Surg. Neural. 12:297~304.
Nagasawa, S., Handa, H., Okumura, A., Naruo, Y., Moritake, K., and Hayashi, K. (1980). Mechanical
properties of human cerebral arteries: Part 2 Vasospasm. Surg. Neural. 14:285~290.
Nagasawa, S., Handa, H., Naruo, Y., Moritake, K., and Hayashi, K. (1982). Experimental cerebral va-
sospasm: Arterial wall mechanics and connective tissue composition. Stroke 13:595--600.
Nagasawa, S., Handa, H., Naruo, Y., Watanabe, H., Moritake, K., and Hayashi, K. (1983). Experimental
cerebral vasospasm: Part 2 contractility of spastic arterial wall. Stroke 14:579~584.
Nerem, R. M. (1992). Vascular fluid mechanics, the arterial wall, and atherosclerosis. ASME J Biomech.
Eng. 114:274-282.
Palmer, R. E., Brady, A. J., and Roos, K. P. (1996). Mechanical measurements from isolated cardiac
myocytes using a pipette attachment system. Am. J Physiol. 270:C697~C704.
Patel, D. J., Greenfield, J. C., and Fry, D. L. (1964). In vivo pressure-length-radius relationship of certain
blood vessels in man and dog. In Attinger, E. 0., ed., Pulsatile Blood Flow. New York: McGraw-Hill.
Peterson, L. H., Jensen, R. E., and Parnell, R. (1960). Mechanical properties of arteries in vivo. Circ. Res.
8:622--639.
Reneman, R. S., van Merode, T., Hick, P., Muytjens, A.M. M., and Hoeks, A. P. G. (1986). Age-related
changes in carotid artery wall properties in men. Ultrasound in Medicine and Biology 12:465--471.
Ricci, D., Tedesco, M., and Grattarola, M. (1997). Mechanical and morphological properties ofliving 3T6
cells probed via scanning force microscopy. Microsc. Res. Tech. 36: 165~ 171.
Richter, H. A., and Mittermayer, C. H. (1984). Volume elasticity, modulus of elasticity and compliance of
normal and atherosclerotic human aorta. Biorheology 21:723~734.
64 K. Hayashi
Ridge, M. D., and Wright, V. (1966). The directional effect of skin - A bioengineering study of skin with
particular reference to Langer's lines. J. Invest. Dermatol. 46:341-346.
Roach, M. R., and Burton, A. C. (1957). The reason for the shape of the distensibility curves of arteries.
Canad. J. Biochem. Physiol. 35:681-690.
Sato, M., Levesque, M. J., and Nerem, R. M. (1987a). An application of the micropipette technique to
the measurement of the mechanical properties of cultured bovine aortic endothelial cells. ASME J.
Biomech. Eng. 109:27-34.
Sato, M., Levesque, M. J., and Nerem, R. M. (1987b). Micropipette aspiration of cultured bovine aortic
endothelial cells exposed to shear stress. Arteriosclerosis 7:276-286.
Shroff, S. G., Saner, D. R., and La!, R. (1995). Dynamic micromechanical properties of cultured rat atrial
myocytes measured by atomic force microscopy. Am. J. Phys. 269:C286-C292.
Takamizawa, K., and Hayashi, K. (1987). Strain energy density function and uniform strain hypothesis for
arterial mechanics. J. Biomech. 20:7-17.
Thoumine, 0., and Ott, A. (1997). Time scale dependent viscoelastic and contractile regimes in fibroblasts
probed by microplate manipulation. J. Cell Sci. 110:2109-2116.
Tong, P., and Fung, Y. C. (1976). The stress-strain relationship for the skin. J. Biomech. 9:649-657.
Vaishnav, R. N., Young, J. T., and Patel, D. J. (1973). Distribution of stresses an strain energy density
through the wall thickness in a canine aortic segment. Circ. Res. 32:577-583.
Vaishnav, R.N., Vossoughi, J., Patel, D. J., Cothran, D. J., Coleman, B. R., and !son-Franklin, E. L. (1990).
Effect of hypertension on elasticity and geometry of aortic tissue from dogs. ASME J. Biomech. Eng.
112:70-74.
Valberg, P. A., and Feldman, H. A. (1987). Magnetic particle motions within living cells. Measurement of
cytoplasmic viscosity and motile activity. Biophys. J. 52:551-561.
Vawter, D. L., Fung, Y. C., and West, J. B. (1978). Elasticity of excised dog lung parenchyma. J. Appl.
Physiol. 45:261-269.
Wang, N., and Ingber, D. E. (1994). Control of cytoskeletal mechanics by extracellular matrix, cell shape,
and mechanical tension. Biophys. J. 66:2181-2189.
Weisenborn, A. L., Khorsandi, M., Kasas, S., Gotzos, V., and Butt, H. J. (1993). Deformation and height
anomaly of soft surfaces studied with an AFM. Nanotechnology 4: 106-113.
Woo, S. L.-Y., Lubock, P., Gomez, M.A., Jemmott, G. F., Kuei, S.C., and Akeson, W. H. (1979). Large
deformation nonhomogeneous and directional properties of articular cartilage. J. Biomech. 12:437-
446.
Yamamoto, E., Hayashi, K., and Yamamoto, N. (1999). Mechanical properties of collagen fascicles from
the rabbit patellar tendon. ASME J. Biomech. Eng. 121: 124-131.
Zahalak, G. 1., McConnaughey, W. B., and Elson, E. L. (1990). Determination of cellular mechanical
properties by cell poking, with an application to leukocytes. ASME J. Biomech. Eng. 112:283-294.
Nonlinear Elasticity, Anisotropy, Material Stability and
Residual Stresses in Soft Tissue
R.W. Ogden
Department of Mathematics, University of Glasgow, Glasgow Gl2 8QW, UK

E-mail: rwo@maths.gla.ac.uk, Home Page: http://www.maths.gla.ac.ukrrwo/
Abstract. In this chapter the basic equations of nonlinear elasticity theory needed for the
analysis of the elastic behaviour of soft tissues are summarized. Particular attention is paid
to characterizing the material symmetries associated with the anisotropy that arises in soft
tissue from its fibrous constituents (collagens) that endow the material with preferred di-
rections. The importance of the issue of convexity in the construction of constitutive laws
(strain-energy functions) for soft tissues is emphasized with reference to material stabil-
ity. The problem of extension and inflation of a thick-walled circular cylindrical tube is
used throughout as an example that is closely associated with arterial wall mechanics. This
is discussed first for isotropic materials, then for cylindrically orthotropic materials. Since
residual stresses have a very important role in, in particular, arterial wall mechanics these are
examined in some detail. Specifically, for the tube extension/inflation problem the residual
stresses arising from the assumption that the circumferential stress is uniform under typical
physiological conditions are calculated for a representative constitutive law and compared
with those calculated using the 'opening angle' method.
1 Nonlinear Elasticity
The mathematical framework for describing the mechanical behaviour of biological soft tissues
has much in common with that used in rubber elasticity, but there are significant differences in
the structures of these materials and in the way that soft tissues and rubber respond under applied
stresses. Figure 1 compares, for example, the typical simple tension stress-stretch response of
rubber (left-hand figure) with that of soft tissue. An important characteristic of soft tissues is the
initial large extension achieved with relatively low levels of stress and the subsequent stiffening
at higher levels of extension, this being associated with the recruitment of collagen fibres as they
become uncrimped and reach their natural lengths, whereupon their significant stiffness comes
into play and overrides that of the underlying matrix material. The distribution of collagen fibres
leads to the pronounced anisotropy in soft tissues, which distinguishes them from the typical
(isotropic) rubber. For a detailed discussion of the morphological structure (histology) of arteries
and relevant references we refer to Humphrey (1995), Holzapfel et al. (2000) and Holzapfel
(2001a, b).
A more detailed picture of the response of soft tissue is illustrated in Figure 2. This shows
results from in vitro experiments on a human iliac artery which is subjected to extension and
inflation. Figure 2(a) shows a plot of the internal pressure against the circumferential stretch >..e
for a series of fixed values of the (reduced) axial load. The curves show the typical stiffening
referred to above. In Figure 2(b) the pressure is plotted against the axial stretch >..z, again for
66 R. W. Ogden
(a) (b)
t
Figure 1. Typical simple tension response of (a) rubber and (b) soft tissue. Nominal stress t ;:::: 0 plotted
against stretch >. 2:: 1.
fixed values of the axial load. An interesting feature here is that a transitional value of the axial
stretch (approximately 1.53) is identified, which is the in vivo value of the axial stretch. This
value is unaffected by changes in pressure and corresponds to an axial load of 0.99 N. For lower
(higher) values of the stretch the pressure stretch curves have a positive (negative) gradient, and
the transitional value may therefore be referred to as an inversion point.
The data in Figure 2 are from an external iliac artery of a 52 year old female without any
cardiovascular risk factors. The artery was healthy in the sense that it was not stenotic. By con-
trast, data from the other specimens considered by Schulze-Bauer et al. (200 1), from individuals
ranging from 57 to 87 years old (with cardiovascularrisk factors), have qualitative characteristics
similar to those shown in Figure 2, but there are some numerical differences. In particular, the
value of the axial stretch at the inversion point is typically between 1.07 and 1.25, significantly
less than the value seen in Figure 2.
Generally, soft tissue material can be regarded as incompressible and anisotropic. The degree
and type of anisotropy depends very much on the tissue considered and its topographic location.
Some tissues (such as tendon and ligament) are transversely isotropic and others (such as arter-
ies) orthotropic, for example. Moreover, their stress-strain response is highly nonlinear, with the
typical rapid stiffening with pressure arising from the recruitment of collagen fibrils and giving
rise to the markedly anisotropic behaviour (Nichols and O'Rourke, 1998), as is illustrated in
Figure 2. In some situations the mechanical response can be treated as purely elastic. For exam-
ple, the passive behaviour of large proximal (close to the heart) arteries such as the aorta, and
the iliac arteries can be regarded as essentially elastic, while the response of distal arteries, on
the other hand, is viscoelastic. These notes, however, are concerned solely with elasticity and its
application to some basic problems and geometries relevant to the characterization of the elastic
response of soft tissues. The theory applies to many different soft tissues, but, for purposes of
illustration, we confine much of the discussion to the analysis of the elasticity of arteries.
In this first section we summarize the basic equations and notation of nonlinear elasticity the-
ory necessary for the continuum description of the mechanical properties of soft tissues, and we
Nonlinear Elasticity, Anisotropy, Material Stability and Residual Stresses in Soft Tissue 67
(a)
35.0
.....
:.
.....
~
30.0
ON
Q;;- 25.0
!
= 20.0
0.70N
~
D. 15.0
0.99N
ii 1.48N
E
s
10.0
....c 5.0
0.0
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3
Circumferential stretch le [-]

(b)
35.0
.....Ia 30.0
ON 1.98N
A.
.....
~ 0.31N 2.47N
Q;;- 25.0
0.70N 2.94N
!
=
.iJ
20.0 0.99N 3.94N
15.0 1.48N 5.90N

CL.
.
ii
c
s
10.0 9.90N
....c 5.0
0.0
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
Axial stretch Az [-]
Figure 2. Typical characteristics of the response of a human iliac artery under pressure and axial load
(Schulze-Bauer et al., 2001). Dependence of the internal pressure H kPa on (a) the circumferential stretch
and (b) the axial stretch at values of the axial load up to 9.9N held constant during the deformation process.
68 R. W. Ogden
examine some illustrative examples of elastic deformations for incompressible isotropic elastic
materials. For a more detailed account we refer to, for example, Ogden (1997, 2001), Holzapfel
(2000) and Fu and Ogden (200 1). See, also, the chapter by Holzapfel (200 1b) in this volume.
1.1 Kinematics
Let X and x, respectively, denote the position vector of a material point in some reference config-
uration, denoted Br, and the (deformed) current configuration, denoted B, which may vary with
time t. The motion (or time-dependent deformation) from Br to B is known when x is specified
as a function of X and t, and we write this in the form
X= X(X, t), (1)
where x is the function describing the motion. For each t, x is invertible and satisfies appropriate
regularity conditions.
The deformation gradient tensor, denoted F, is given by
F = Gradx (2)
and has Cartesian components Fia = 8xd8Xa. where Grad is the gradient operator in Brand
Xi and X 01 are the components ofx and X, respectively, i, a E {1, 2, 3}. Local invertibility of the
deformation requires that F be non-singular, and the usual convention that
J = detF > 0 (3)
is adopted, wherein the notation J is defined.

The velocity v and acceleration a of a material particle are given, respectively, by
ax a2x
v = 8t(X,t), a= ()t 2 (X, t), (4)
these being the first and second material time derivatives of X.

The (unique) polar decompositions
F=RU=VR, (5)
then follow, R being a proper orthogonal tensor and U, V positive definite and symmetric ten-
sors (the right and left stretch tensors, respectively). The tensors U and V have the spectral
decompositions
L L
3 3
U = A.iu(i) 129 u(i), V = A.iv(i) 129 v(il, (6)
i=l i=l
respectively, where Ai > 0, i E {1, 2, 3}, are the principal stretches, u(il are the (unit) eigen-
vectors ofU, called the Lagrangian principal axes, y(i), the (unit) eigenvectors ofV, called the
Eulerian principal axes, and 129 denotes the tensor product.
The left and right Cauchy-Green deformation tensors, denoted respectively by Band C, are
defined by
B = FFT V 2 , = (7)
and the principal invariants of B (equivalently of C) are defined by
h = tr(B), h =
1
2[11
2
-
2
tr(B )], h = det(B) =(detF) 2• (8)
The Green strain tensor, defined by
1
E= 2(c- I), (9)
where I is the identity tensor, will be required in the following sections.

If the material is incompressible, which is usually taken to be the case both for rubber and
for soft tissues, then the incompressibility constraint
h = 1 (10)
must be satisfied.
1.2 Stress and the equations of motion
The equation of motion may be expressed in the form
Div s = Pr a = Pr X,tt' (11)
where Sis the nominal stress tensor (the transpose ofthe first Piola-Kirchhoffstress tensor), Pr
is the mass density of the material in Br and ,t signifies the material time derivative. Body forces
have been omitted. Equation (11) is the Lagrangian version of the equation of motion, with (X, t)
used as the independent variables.
The global counterpart of the local balance equation ( 11) may be written
(12)
where N is the unit outward normal to the boundary 8Br of Br, dAis the area element on 8Br
and dV the volume element in Br. This serves to identify the traction vector STN per unit area
of 8Br (also referred to as the load or stress vector).
The equation of motion (11) may, equivalently, be written in the Eulerian form
div£T = pa, (13)
where the symmetric tensor £T is the Cauchy stress tensor, div is the divergence operator with
respect to x and p is the material density in B, with (x, t) as the independent variables and a
treated, through the inverse of ( 1), as a function of x and t.
The two densities are connected through
Pr = pJ, (14)
so that if the material is incompressible then p = Pr·

70 R. W. Ogden
The nominal and Cauchy stress tensors are related by
(15)
The Biot stress tensor, denoted T, and the second Fiola-Kirchhoff stress tensor, denoted T,
which are both symmetric, will also be used in what follows. They are defined by
(16)
and connected through

1
T = 2(rU + Ur). (17)
1.3 Hyperelasticity
We consider an elastic material for which the material properties are characterized in terms of
a strain-energy function (per unit volume), denoted W = W (F) and defined on the space of
deformation gradients. This theory is known as hyperelasticity. For an inhomogeneous material,
i.e. one whose properties vary from point to point, W depends on X in addition to F, but we do
not indicate this dependence explicitly in what follows.
For an unconstrained hyperelastic material the nominal stress is given by
aw
S = H(F) =: aF, (18)
wherein the notation H is defined. The tensor function H is referred to as the response function
of the material relative to the configuration Br in respect of the nominal stress tensor. In com-
ponents, the derivative in (18) is written sai = awI aFia' which provides our convention for
ordering of the indices in the partial derivative with respect to F.
For an incompressible material the counterpart of (18) is
aw _1
S = 8F - pF ' detF = 1, (19)
where pis the Lagrange multiplier associated with the incompressibility constraint and is referred
to as the arbitrary hydrostatic pressure.
The Cauchy stress tensor corresponding to ( 18), on use of ( 15), is then seen to be given by
__ 1 aw
u = G(F) = J F BF , (20)
wherein the response function G associated with u is defined. As for H, the form ofG depends on
the choice of reference configuration, and G is referred to as the response function of the material
relative to Br associated with the Cauchy stress tensor. Unlike H, however, G is a symmetric
tensor-valued function. For incompressible materials (20) is replaced by
aw
u = F BF -pi, det F = 1. (21)
If the configuration Br is stress free then it is referred to as a natural corifiguration. Here, we

take Wand the stress to vanish in Br, so that, for an unconstrained material,
aw
W(I) = 0, 8F (I)= 0, (22)
with appropriate modifications in the case of an incompressible material. If the stress does not
vanish in Br then this configuration is said to be residually stressed. In such a configuration the
traction must vanish at all points of the boundary, so that, a fortiori, residual stress is inhomoge-
neous in character (i.e., it cannot be uniform). Residual stresses are important in the context of
biological tissues and the restriction (22) will be removed later, in Section 4, where we consider
the consequences of residual stress for the constitutive law and response of the material.
Objectivity The elastic stored energy is required to be independent of superimposed rigid mo-
tions of the form
x* = Q(t)x + c(t), (23)
where Q is a proper orthogonal (rotation) tensor c is a translation vector. The resulting deforma-
tion gradient is QF and it therefore follows that
W(QF) = W(F) (24)
for all rotations Q. A strain-energy function satisfying this requirement is said to be objective.
Use of the polar decomposition (5) and the choice Q = RT in (24) shows that
W(F) = W(U). (25)
Thus, W depends on F only through the stretch tensor U and may therefore be defined on
the class of positive definite symmetric tensors. Equivalently, through (7) and (9), W may be
regarded as a function ofthe Green strain E.
Expressions analogous to ( 18) and ( 19) can therefore be written down for the Biot and second
Piola-Kirchhoff stresses. Thus,
aw aw
T= 8U' T= 8E' (26)
and
aw _1 aw _1
T = au - pU , det U = 1, T = BE - p C , det C = 1 (27)
for unconstrained and incompressible materials respectively, where C = I + 2E. Note that
when expressed as a function of U or E the strain energy automatically satisfies the objectivity
requirement.
Material symmetry Mathematically, there is no restriction so far other than (22) and (24) on
the form that the function W may take. However, the predicted stress-strain behaviour based on
the form of W must on the one hand be acceptable for the description of the elastic behaviour of
real materials and on the other hand make mathematical sense.
Further restrictions on the form of W arise if the material possesses symmetries in the con-
figuration Br. Material symmetry (relative to a given reference configuration) is identified by
72 R. W. Ogden
transformations of the reference configuration that do not affect the material response. Consider
a change from the reference configuration Br (in which material points are identified by position
vectors X) to a new reference configuration B~ (with material points identified by X'), let F'
denote the deformation gradient relative to B~ and let Grad X' be denoted by P. If the material
response is unchanged then
W(F'P) = W(F') (28)
for all deformation gradients F'. This states that the strain-energy function is unaffected by a
change of reference configuration with deformation gradient P. The collection of P for which
(28) holds forms a group, which is called the symmetry group of the material relative to Br. As
already mentioned biological soft tissues are distinguished, in particular, by the anisotropy of
their structure. We will examine the appropriate type of anisotropy in Section 2, but initially we
shall, for simplicity, focus on the development of the theory in the case of isotropy.
Isotropy To be specific we now consider isotropic elastic materials, for which the symmetry
group is the proper orthogonal group. Then, we have
W(FQ) = W(F) (29)
for all rotations Q. Bearing in mind that the Q's appearing in (24) and (29) are independent the
combination of these two equations yields
W(QUQT) = W(U) (30)
for all rotations Q, or, equivalently, W(QVQT) = W(V). Equation (30) states that W is an
isotropic function ofV. It follows from the spectral decomposition (5) that W depends on U only
through the principal stretches ,\ 1 , ,\2 , ,\ 3 . To avoid introducing additional notation we express
this dependence as W (,\ 1 , ,\ 2 , ,\ 3 ); by selecting appropriate values for Q in (30) we may deduce
that W depends symmetrically on -\1, ,\ 2 , ,\3, i.e.
(31)
A consequence of isotropy is that the Biot stress T is coaxial with U and, equivalently, the
Cauchy stress G" is coaxial with V. Hence, in parallel with (6), we have
3 3
T = 2::= tiu(i) ® u(i), (T = 2::= O"iV(i) @ y(i)' (32)
i=l i=l
where ti, are the principal Biot stresses and O"i are the principal Cauchy stresses. For an uncon-
strained material,
(33)
while for an incompressible material these are replaced by
aw _1
t; = a,\i - P\ ' (34)
Note that in (33) and (34) there is no summation over the repeated index i.
For an isotropic material the symmetric dependence (31) of W on the principal stretches is
equivalent to W being regarded as a function of the (symmetric) principal invariants I 1 , h, Is
defined by (8). In terms of the invariants h, I 2 , Is the Cauchy stress tensor for an unconstrained
isotropic elastic material may be written
(35)
where the coefficients a 0 , a 1 , a 2 are functions of I 1 , h, Is given by
a1 -
_ 2I -1;z(aw
3
aw)
ah + h a[z , (36)
with W is now regarded as a function of h, lz, Is. For an incompressible material the corre-
sponding expression is
(37)
where pis the arbitrary hydrostatic pressure, a 1 and a 2 are again given by (36) (but with I 3 = 1),
and W is now regarded as a function of h and I 2 alone.
1.4 Initial-boundary-value problems
For an unconstrained material we now consider the equation of motion (11) together with the
stress-deformation relation (18), and the deformation gradient (2) coupled with (1 ). Thus,
Div (~~) = PrX,tt, F = Gradx, X= x(X, t), (38)
An initial-boundary-value problem is obtained by supplementing (38) with appropriate boundary

and initial conditions.
Typical boundary conditions arising in problems of nonlinear elasticity are those in which x
is specified on part of the boundary, as; c asr say, and the stress vector on the remainder, as:,
so that as; U as: = asr and as; n as: = 0. We write
x =~(X, t) on as~, (39)

STN = s(F, X, t) on as:, (40)
where ~ and s are specified functions. In general, s may depend on the deformation and this is
indicated in (40) by the explicit dependence of s on the deformation gradient F. If, for example,
the boundary traction in (40) is associated with a hydrostatic pressure, P say, so that un = - Pn,
where n is the unit outward normal to as, then s depends on the deformation in the form
(41)
We record here that nand N are related through Nanson's formula nda = JF-TNdA, where
dais the area element on as. In the important special case in which the surface traction defined
by (40) is independent ofF and t it is referred to as a dead-load traction.
74 R. W. Ogden
A variety of possible initial conditions may be specified. A typical set corresponds to pre-
scription of the initial values ofx and x,t· Thus,
x(X, 0) = Xo(X), (42)
where xo and vo are prescribed functions. A basic initial-boundary-value problem of nonlinear

elasticity is then characterized by (38)-(40) with (42).
In components, the equation of motion in (38) can be written
fJ2x
AaiJ)j aXaO~J) = Pr Xi,tt, (43)
fori E {1, 2, 3}, where the coefficients AaiJ3j are defined by

fJ2W
Aaii3i = Ai3Jai = aF;aaFii3 (44)
The pairwise symmetry of the indices in (44) should be noted.

For incompressible materials the corresponding equations, obtained by substituting (19)1
into (11 ), are
(45)
subject to (10), where the coefficients are again given by (44).

The coefficients AaiJ3j are, in general, nonlinear functions of the components of the deforma-
tion gradient. Explicit expressions for the components Aai/3j in respect of an isotropic material
can be found, for example, in Ogden ( 1997), and we shall make use of them in Section 3. We
emphasize that in the above Cartesian coordinates are being used. In Section 1.5.2 and subse-
quently we shall require the radial equation in cylindrical polar coordinates. This will be given
at the appropriate point. More general expressions for the (equilibrium) equations in cylindrical
polar coordinates and other coordinate systems are given in Ogden ( 1997), for example.
In order to analyze such initial-boundary-value problems additional information about the na-
ture of the function W is required. This information may come from the construction of special
forms of strain-energy function based on comparison of theory with experiment for particular
materials, it may arise naturally in the course of solution of particular problems, from consider-
ations of stability for example, or may be derived from mathematical requirements on the prop-
erties that W should possess in order for existence of solutions to be guaranteed, for example.
Consideration of the form of W is therefore absolutely crucial, and some aspects of this will be
examined in the course of the following sections.
Finally in this section, we give an expression for the coefficients AaiJ3j in terms of the deriva-
tives of W with respect to the Green strain E since the latter (or, equivalently, C) is often used in
computational formulations of the equations of motion (or equilibrium). This is
82 W oW
Aai;3j = KrFjo 8Ea,8E)3o + 8;j 8Eai3. (46)
Note that for an incompressible material the form of these coefficients depends on the point
at which the incompressibility condition is invoked in calculating the partial derivatives. This
does not lead to any inconsistencies since differences are absorbed by the arbitrary hydrostatic
pressure term. Expressions analogous to (46) may be written down for other strain or deformation
measures such as U but we do not require them here.
1.5 Examples
Homogeneous deformations We consider first an elementary time-independent problem in

which the deformation is homogeneous, so that the deformation gradient F is constant. Specifi-
cally, we consider the pure homogeneous strain defined by
(47)
where the principal stretches >. 1 , >. 2, >. 3 are constants. For this deformation F = U = V, R = I
and the principal axes of the deformation coincide with the Cartesian coordinate directions and
are fixed as the values of the stretches change. Thus, F = diag( >. 1, >. 2, >. 3 ). For an unconstrained
isotropic elastic material the associated principal Biot stresses are given by (33)1. These equa-
tions serve as a basis for determining the form of W from triaxial experimental tests in which
>.1, >.2, A3 and t 1, t 2, t 3 are measured. If biaxial tests are conducted on a thin sheet of material
which lies in the (X1,X2)-plane with no force applied to the faces of the sheet then equations
(33)1 reduce to
(48)
and the third equation gives >. 3 implicitly in terms of >. 1 and >. 2 when W is known.
The biaxial test is particularly important when the incompressibility constraint
(49)
holds since then only two stretches can be varied independently and biaxial tests are sufficient
to obtain a characterization of W. The counterpart of (33) for the incompressible case is given
by (34 ). It is convenient to make use of (49) to express the strain energy as a function of two
independent stretches, and for this purpose we define
(50)
This enables p to be eliminated from equations (34) and leads to
(51)
It is important to note that, because of the incompressibility constraint, equation (51) is unaffected
by the superposition of an arbitrary hydrostatic stress. Thus, without loss of generality, we may
set a3 = 0 in (51). In terms of the principal Biot stresses we then have simply
aw (52)
h = 8>.1'
76 R. W. Ogden
which provides two equations relating >. 1, >.2 and t 1, t 2 and therefore a basis for characterizing
W from measured biaxial data. This situation is special for an isotropic material, and, as we shall
see in Section 2, biaxial tests alone are not sufficient for the characterization of the strain-energy
function of an anisotropic material.
There are several special cases of the biaxial test which are of interest, but we just give the
details for one, namely simple tension, for which we set t 2 = 0. By symmetry, the incompress-
ibility constraint then yields >. 2 = >. 3 = >.~ 112 . The strain energy may now be treated as a
function of just >. 1, which we write as >., and we define
(53)
and (52)1 reduces to

t = w'(>.), (54)
where the prime indicates differentiation with respect to >. and t 1 has been replaced by t. It is the
notation(>., t) that has been used in Figure 1.
For future reference we note here the form of (51) when W is regarded as a function of I 1
and h. Thus,
-2 -2 4 2 (aw
a1- a3 = 2>.1 >.2 (>.1>.2 -1) 8h + >.2 8h '
2aw) (55)
a2 - a3
-2 -2 2 4
= 2>.1 >-2 (>.1 >-2 - 1)
(aw
ah + >-12aw)
ah , (56)
where W(h, h)= W(>.1, >.2).
Extension and inflation of a thick-walled tube Here we examine a prototype example of a non-
homogeneous deformation that is particularly relevant to the mechanics of arteries. We consider
a thick-walled circular cylindrical tube whose initial geometry is defined by
A :S R :S B, 0 :S 8 :S 21r, 0 :S Z :S L, (57)
where A, B, L are positive constants and R, 8, Z are cylindrical polar coordinates. The tube is
deformed so that the circular cylindrical shape is maintained, and the material of the tube is taken
to be incompressible. The resulting deformation is then described by the equations
(58)
where r, B, z are cylindrical polar coordinates in the deformed configuration, >-z is the (uniform)
axial stretch and a is the internal radius of the deformed tube.
The principal stretches >. 1, >. 2 , >. 3 are associated respectively with the radial, azimuthal and
axial directions and are written
(59)
wherein the notation>. is introduced. It follows from (58) and (59) that
2 R2 2 B2 2
AaAz- 1 = A 2 (>. Az- 1) = A 2 (>.b>.z- 1), (60)
where
.Xa =a/A, .Xb = b/B. (61)
Note that for given Az and A/ B, Ab depends on Aa. For a fixed value of Az the inequalities
(62)
holdduringinflationofthetube, withequalityholdingifandonlyifA = X;- 1/ 2 for A :S R :S B.

We use the notation (50) for the strain energy but with .X 2 = .X and Ag = Az as the indepen-
dent stretches, so that
(63)
ow ow
Hence
CJ2 - CJ1 = .A o.X , (Jg - CJ1 = Az o.Az. (64)
Note that locally the deformation has the same structure as that of the pure homogeneous strain
discussed in Section 1.5 .1.
When there is no time dependence the equation of motion (13) reduces to the equilibrium
equation div u = 0. For the considered symmetry its cylindrical polar component form reduces
to the single scalar equation
dCJrr 1
-d- + -(CJrr- CJee) = 0, (65)
r r
where CJrr = CJ 1, CJee = CJ 2 in terms of the principal Cauchy stresses. This is to be solved in
conjunction with the boundary conditions
-PonR= A
(J rr ={0 on R = B (66)
corresponding to pressure P (2': 0) on the inside of the tube and zero traction on the outside.
On use of (58)-( 61) the independent variable may be changed from r to .X, and integration of
(65) and use of the boundary conditions (66) then yields
A A
p = { a (.X2 Az- 1)-1 oW d.X. (67)

JAb OA
Since, from (60), Ab depends on Aa, equation (67) provides an expression for Pas a function of
Aa (equivalently of the internal radius) when Az is fixed. In order to hold Az fixed an axial load,
N say, must be applied to the ends of the tube. This is given by
A A A
Nj1rA 2 = (.X~.Az- 1) { a (.X 2 Az- 1)- 2 ( 2.Az ~~ -.X 0 ~)-Xd.X + P.X~.

0 (68)
JAb z
Equation (68) applies to a tube with closed ends so that the pressure contributes to the axial load.
An illustrative plot of the pressure P calculated from (67) for a fixed value of Az (= 1.2)
is shown in Figure 3(a) in dimensionless form in respect of the simple incompressible isotropic
strain-energy function with W given by
W(.A, Az) = 2 ~ (.An+ A~+ A-n .A_;-n- 3), (69)

n
78 R. W. Ogden
(a) (b)
P* P*
0.95 1.1 1.2
Figure 3. Plot of the dimensionless pressure P* against stretch Aa for different wall thicknesses and an
axial pre-stretch Az = 1.2 in respect of (a) the strain-energy function (69) with n = 24, and (b) a typical
rubberlike material.
where the constant IL is the shear modulus of the material and n is a dimensionless material con-
stant. This is a special case ofthe class of strain-energy functions introduced by Ogden (1972).
The dimensionless pressure is given by P* = nP121" and results are shown for n = 24. In the fig-
ure results are compared for different values of the ratio A 2 I B 2 , specifically 0.4, 0.63, 0.77, 0.85,
reading from left to right in the figure. The feature to be noted here is the increasing stiffuess with
wall thickness. The results are contrasted with those for a typical rubberlike material shown in
Figure 3(b) for the same values of A 2 I B 2 (reading from top to bottom). The non-uniqueness
in the relationship between P* and >-a evident in Figure 3(b) may be (but is not necessarily)
associated with loss of stability of the symmetrical configuration and the possible emergence of
bulges in the tube wall, a phenomenon which is excluded when P* is a monotonically increas-
ing function of >.a, as is the situation in Figure 3(a). A detailed treatment of the inflation and
extension of thick- and thin-walled tubes of rubber, with particular reference to loss of stability
is given by Haughton and Ogden (1979a, b). For an account of constitutive laws for rubberlike
solids we refer to Ogden (1982), while for a review that includes discussion of both elastomers
and biological tissues the paper by Beatty (1987) is recommended.
2 Anisotropy
2.1 Fibre-reinforced materials
An important characteristic of soft tissues is the anisotropy in their mechanical properties that
arises as a result of tensile stresses. This is associated with the collagen fibres that are distributed
in the material and endow the material with directional properties. In some soft tissues (for ex-
ample, ligament and tendon) this leads to the material having, in a macroscopic sense, a single
preferred direction (see, for example, the discussion in Holzapfel, 200 la). The properties of the
material can therefore be regarded as transversely isotropic. Other soft tissues, such as artery
walls have two distinct distributions of collagen fibre directions and these can be associated
with two preferred directions. Here, we illustrate the structure of the strain-energy function of
an anisotropic elastic solid both for transverse isotropy and for its extension to the case of two
preferred directions.
One family of fibres: transverse isotropy Firstly, we consider transverse isotropy. Let the unit
vector M be a preferred direction in the reference configuration Br of the material. The material
response is then indifferent to arbitrary rotations about the direction M and by replacement of
M by - M. Such a material can be characterized with a strain energy which depends on F and
the tensor M 0 M, as described by Spencer (1972, 1984); see, also, Holzapfel (2000). Thus, we
write W (F, M 0 M) and, for an unconstrained material, the required symmetry reduces W to
dependence on five invariants, namely I 1 , h, I 3 , as defined by (8), and the additional invariants
l4 and l5 depending on M and defined by
(70)
where C is the right Cauchy-Green deformation tensor defined in (7). We note that I 4 has a direct
kinematical interpretation since I~/ 2 represents the stretch in the direction M, but that there is no
similar simple interpretation for Is in general.
On use of ( 18) the resulting nominal stress tensor is expressed in the form
S = 2W1FT + 2W2(I1I- C)FT + 2hW3F- 1 + 2W4M 0 FM

+ 2W5(M 0 FCM + CM 0 FM), (71)
where wi aw a
= 0 ° 0 ' w4
I h i = 1' 5. For an isotropic material the terms in and are omit- w5
ted. Equation (71) describes the response of a fibre-reinforced material with the fibre direction
corresponding to M locally in the reference configuration. The vector field M can be thought of
=
as a field that models the fibres as a continuous distribution.
For an incompressible material h 1, and the counterpart of(71) is
T
S = -pF -1 + 2W1F
- -
+ 2W2(hl- C)F T + 2W4M
"
0 FM
+ 2W5(M 0 FCM + CM 0 FM), (72)
where, extending the notation defined following (56), the notation W is used when W is regarded
as a function of (h, h, l4 ,[5 ). The corresponding Cauchy stress tensor is given by
v v 2 v
u =-pi+ 2W1B + 2W2(hB- B)+ 2W4FM 0 FM

+ 2W5(FM 0 BFM + BFM 0 FM), (73)
where B is the left Cauchy-Green deformation tensor. The symmetry of u can be seen immedi-
ately in (73), and we note that (73) reduces to (37) when the dependence on I 4 and I 5 is omitted.
For the pure homogeneous strain defined by (47) it is interesting to contrast the results for an
isotropic material with those derived from (73). With respect to Cartesian axes, suppose that M
80 R. W. Ogden
has components (cos <p, sin <p, 0), i.e. M lies in the principal plane associated with the stretches
.\ 1 and .\ 2 . Then, we have
(74)
and the connection

(75)
is obtained in this case.
The components of cr are given by
Note that u 12 vanishes if the preferred direction is along one of the coordinate axes.
From (76), (77) and (79) it follows that
(80)
(81)
which extends the formulas (55) and (56) to the case of transverse isotropy.
Because of the incompressibility condition (49), h, h ,14 ,15 , and hence the strain-energy
function, depend only on .\ 1 , .\ 2 and the angle <p. In parallel with the definition (50) we therefore
use the notation
(82)
to indicate this dependence. Note, however, that in general, in contrast to the isotropic situation,
W(.\ 1 , .\ 2 , <p) is not symmetric in .\ 1 and .\2 . It follows that (80) and (81) may be written simply
in the form
(83)
These equations are very similar to equations (51), but here u 11 and u 2 2 are not principal stresses
since the shear stress u 12 does not in general vanish (the principal axes of cr do not coincide
with the principal axes of B, in contrast to the case for isotropy), and, it should be emphasized,
that W(.\ 1 , .\ 2 , <p) is not symmetric in .\ 1 and .\ 2 . The shear stress u12 is required to maintain
the prescribed deformation. From (78) we see that the latter depends on the material properties
through the dependence ofW on ! 4 and Is and (separately) on the value of <p. It is interesting to
note that if <pis treated as a parameter then u 12 can be expressed in the form
.\1.\2 8W (84)
0"12 = .\~ - .\i 8<p .
It follows that, unlike the situation for incompressible isotropic materials, homogeneous biax-
ial deformations are not on their own sufficient to characterize the properties of a transversely
isotropic elastic material.
Let m denote the unit vector in the direction FM. For the considered pure strain FM has
components (.A 1 cos cp, .A 2 sin cp, 0). Suppose that m has components (cos cp*, sin cp*, 0). Then it
follows that the angle cp* is given by
(85)
Thus, in general, the orientation of the fibres is changed by the deformation.
Two families of fibres: orthotropy When there are two families of fibres corresponding to two
preferred directions in the reference configuration, M and M' say, then, in addition to (8) and
(70), the strain energy depends on the invariants
I6 = M' · (CM'), h = M' · (C 2M'), Is= M · (CM'), (86)
and also on M · M' (which does not depend on the deformation); see Spencer (1972, 1984)
for details. Note that Is involves interaction between the two preferred directions, but the term
M · (C 2 M'), which might be expected to appear in the list (86), is omitted since it depends on
the other invariants and on M · M'.
For a compressible material the nominal stress (71) is now extended to
S = 2W1FT + 2W2(I1I- C)FT + 2hW3F- 1 + 2W4M c>9 FM

+ 2W5 (M c>9 FCM + CM c>9 FM) + 2W6M' c>9 FM'
+ 2W7 (M' c>9 FCM' + CM' c>9 FM')
+ Ws (M c>9 FM' + M' c>9 FM), (87)
where the notation Wi = 8Wj8Ii now applies fori = 1, ... , 8. The counterpart of (87) for an
incompressible material may be obtained on the same basis as for (72).
We now use the notation W to represent W for an incompressible material when regarded as
a function of h, h, I 4 , h, I 6 , h, Is. The Cauchy stress tensor is then written
- - 2 -
u =-pi+ 2W1B + 2W2(hB- B ) + 2W4FM c>9 FM
+ 2W5 (FM c>9 BFM + BFM c>9 FM) + 2W6 FM' c>9 FM'
+ 2W1(FM' c>9 BFM' + BFM' c>9 FM')
+ Ws (FM c>9 FM' + FM' c>9 FM), (88)
the component form of which will be used in the following analysis.

For the solution of an initial-boundary-value problem on the basis of equations (38)-( 42), for
example, the form of the stress given by (87) or (88) is required. Clearly, these forms are very
complicated and, in general, little can be achieved without the use of numerical computation, and
this requires an appropriate choice of a particular form of W. A number of specific forms based
on the use of the invariants discussed above have been used in the literature (see, for example,
Humphrey, 1995, 1999, and Holzapfel et al., 2000, for references). Other forms of anisotropic
82 R. W. Ogden
energy function not based directly on the use of these invariants have also been used in the
biomechanics literature, as discussed in Humphrey (1995) and Holzapfel et al. (2000). Here, our
intention is examine two specific and very simple deformations, namely the pure homogeneous
strain and extension/inflation of a tube discussed (for an incompressible isotropic elastic material)
in Sections 1.5.1 and 1.5.2 respectively, in respect of a general form of strain-energy function of
the considered class in order to extract some qualitative and quantitative information about the
nature of the energy function and its predictions.
Pure homogeneous strain. Again we consider the pure homogeneous strain defined by (4 7) and
now we include two fibre directions, symmetrically disposed in the (X 1 , X 2 ) -plane and given by
M = cos<pe 1 + sin<pe2, M' = cos<pe 1 - sin<pe 2, (89)
where the angle <p is constant and e 1 , e 2 denote the Cartesian coordinate directions. Let the
corresponding unit vectors in the deformed configuration be denoted
m I = cos <p * e1 - sm
. <p * e2, (90)
with <p* again being given by (83). The deformation and fibre directions are depicted in Figure
4.
(a)
M'
Figure 4. Pure homogeneous strain of a thin sheet of material with two in-plane symmetrically disposed
families of fibres: (a) undeformed configuration; (b) deformed configuration.
When expressed in terms of ,\ 1 and ..\2 the invariants h, l2 are given by

I ,2 ,2 ,_2,-2 I ,-2 ,-2 ,2,2
1 = "'1 + "'2 + "'1"'2 ' 2 = "'1 + "'2 + "'1 "'2' (91)
while the other invariants are explicitly
I4 = h = .Ai cos 2 'P + .A~ sin 2 'f!, I5 = I7 = \4

At cos 'P
2
+ \42 sm
. 2
A '{!, (92)
Is = .Ai cos 2 'P- .A~ sin 2 'P· (93)

From (88) the components of u are found to be
'2 2 4 2 2
au = -p + 2Wt.At + 2W2(lt.At -.At)+ 2(W4 + W6 + Ws).At cos 'P
y y y y
+ 4(W5 + w1 ).At cos 'f!,

y 4 2
y
(94)
2 2 4 2·2
a22 = -p + 2Wt.A 2 + 2W2(!t.A 2 - .A 2) + 2(W4 + W6- Ws).A 2 sm 'P
y y y y y
4 2
+ 4(W5 + W1 ).A 2 sin 'f!, (95)
y y
at2 = 2[W4- w6 + (W5- W1)(.At + .A2)].At.A2 sm'Pcos'P,

y y y y 2 2 .
(96)
2 2 4
a33 = -p + 2Wt.A 3 + 2W2(lt.A 3 - .A 3), at3 = a23 = 0, (97)
y y
and we note that (97) is identical ifform to (79) but is different in content since W now depends
on I6, h,Is.
In general, since at 2 -::j:. 0, shear stresses are required to maintain the pure homogeneous
deformation and the principal axes of stress do not coincide with the Cartesian axes. However, in
the special case in which the two families offibres are mechanically equivalent the strain energy
must be symmetric with respect to interchange of I4 and I 6 and of I 5 and h. Since, for the
considered deformation, we have I4 = h,h =hit follows that = = w4 and w6, w5 w7
hence, from (96), that at 2 = 0. The principal axes of stress then coincide with the Cartesian axes
and au, a 22 , 0"33 are just the principal Cauchy stresses ITt, a 2, a 3 .
In view of (91) and the fact that I4, I 5 , Is depend on At, .A 2 and 'f!, we may regard the strain
energy as a function of At, .A2 and 'P· We write W(.At, .A2, 'f!), but it should be emphasized that,
as for the transversely isotropic case and unlike for isotropic materials, W is not symmetric with
respect to interchange of any pair of the stretches. Extending the definition (82) we have
W(.A1, .A2, 'P) = W(I1, I2, hIs, Is, I7, Is), (98)
with (91)-(93), and it is straightforward to check that
(99)
which are identical inform to equations (51) except that here W depends on 'P and is not (in
general) symmetric in (.At, .A 2). These equations describe an orthotropic material with the axes
of orthotropy coinciding with the Cartesian axes.
Extension and inflation of a thick-walled tube. For the extension and inflation of a thick-walled
tube the deformation was examined in Section 1.5.2. Since this deformation is locally a pure
homogeneous strain the results discussed for an isotropic material carry over to the considered
anisotropic material with the fibre directions M and M' locally in the ( 8, Z)-plane symmetrically
disposed with respect to the axial direction. The cylindrical polar directions are then the principal
directions of strain (and stress) and the strain energy may be written in the form
(100)
84 R. W. Ogden
where, as in Section 1.5.2, >. = >. 2 and Az = >. 3 respectively are the azimuthal and axial stretches.
Furthermore, the formulas (67) and (68) again apply. It is worth emphasizing that they remain
valid if the fibre directions depend on the radius, i.e. if <p depends on R, and, in particular, if
there are two or more concentric layers with different values of <p. For convenience, we repeat
equation (67) here as
(101)
with the arguments of W made explicit.

It is a straightforward matter to evaluate the integral in ( 101) for particular choices of energy
function, as was illustrated in the case of isotropy in Section 1.5.2. As shown in Figure 3 the
qualitative character of the results is essentially independent of the tube wall thickness and suf-
ficient information concerning the dependence of the pressure-stretch response on the degree of
anisotropy can therefore be determined by considering the thin-wall (membrane) approximation
of(lOl). This has the form
P=EA
-1 -law
>.z a>. (>.,.Az.<p), (102)
where E = HI A, H = B - A being the wall thickness in the reference configuration, and >.
represents any value of the azimuthal stretch through the wall (such values differ only by a term
of order E).
For definiteness we consider an energy function that is a natural extension of (69) to the type
of anisotropy considered here. This has the form
W(>., A2 , <p) = [Jh (<p)(An- 1- n In .A)+ jL2(<p)(A~- 1- n In Az)

+ p3(>.-n x;n -1 + nln(>.>.z))Jin, (103)
where the logarithmic terms are needed to ensure that the stresses vanish in the undeformed
configuration, p 3 is a material constant and PI (<p) and P2 (<p) are material parameters dependent
of the angle <p. Equation (1 03) is a special case of a form of energy function used by Ogden and
Schulze-Bauer (2000). Note that (69) is recovered by setting PI = P2 = P3 = 2pln.
In respect of (103) equation (1 02) gives, in dimensionless form,
p* =A
- z
PIE ,..3
II = 11•
t'"'l
An-2 _ (t'"'l
11• _ 1)>. -2 _A -n-2 A-n
z ' (104)
where Pi = p 1 I p 3. Note that ( 104) is independent of P2. The results for isotropy are recovered
by setting Pi = 1. The material can be regarded as reinforced (weakened) in the circumferential
direction (relative to the radial direction) if Pi > 1 (pi < 1). Results for Pi = 0.5, 1, 2 are
plotted in Figure 5 for comparison, with Az set to the value 1.2, as for Figure 3. We recall that
for isotropy the inequality >. 2 Az 2: 1 must hold for inflation following an initial axial stretch.
For an anisotropic material this must be replaced by an inequality on >. n whose lower limit is
determined by setting P = 0 in (104). This is reflected in the curves in Figure 5, which cut
the >. axis at different points. The upper, middle and lower curves in Figure 5 correspond to
Pi = 2, 1, 0.5 respectively. For illustrative purposes only the value n = 10 has been used for the
above calculations.
Figure 5. Plot of the dimensionless pressure P* against the azimuthal stretch A for fixed Az and for values
J-ti = 2, 1, 0.5 of the anisotropy parameter, corresponding to the upper, middle and lower curves respec-
tively.
The membrane counterpart of ( 102) for equation (68) has the form
(105)
where F is interpreted as the reduced force on the ends of the tube. With F = 0 the equivalent
of equations (102) and (105) (in different notation) were solved numerically by Holzapfel and
Gasser (200 1) in respect of an energy function containing a term exponential in I 4 to obtain
plots of the pressure as a function of,\ and (separately and equivalently) of Az. This reveals some
interesting characteristics. For a certain range of values of cp the pressure increases monotonically
with ,\ but for other values it is non-monotonic and,\ decreases initially as the pressure increases
before ultimately reaching a minimum and then increasing. Similarly for the dependence of P
on Az. This so-called inversion effect is a reflection of the inversion seen in the actual data for
arteries in Figure 2(b). The turning points are determined by solution of the equation
in conjunction with (1 05) for constant F* = F jurA 2 , as pointed out by Ogden and Schulze-
Hauer (2000). Equation (106) is obtained from (102) and (105) by setting d-\z/dP = 0 at con-
stant F. In order to predict the inversion effect using an energy function of the type ( 103) it needs
to be extended to include two or more terms with different values of n, but we omit details of
this here. It should be pointed out that this inversion effect is not a feature specific to anisotropy
since it may be found for particular choices of isotropic energy function.
Although the membrane approximation gives a good qualitative picture of the pressure-
stretch behaviour it should be used with caution. For example, membrane theory is not able
to account for the through-thickness stress distribution in arterial walls or the important influ-
ence of residual stresses that are present in arterial wall layers. To account for these influences it
86 R. W. Ogden
is necessary to use a 'thick-wall' model. Both residual stresses and stress distributions through
the wall thickness will be discussed in Section 4.
3 Convexity and Material Stability
The concepts of convexity and material stability are closely related and have important roles
to play in the construction of constitutive laws for soft biological tissues and for the analysis
of boundary-value problems. In this section we examine some connections between these no-
tions and, in particular, between the convexity of contours of constant strain energy and local
material stability. One point to emphasize is that these notions depend on the choice of strain
or deformation measure and, for purposes of illustration, we begin, in the following section, by
considering constant energy contours for two specific forms of energy function, one isotropic
and one anisotropic.
3.1 Contours of constant energy
In the notation W that we have used here to represent the strain-energy function for an incom-
pressible material when regarded as a function of the invariants, the (isotropic) strain-energy
function introduced by Delfino et al. (1997) is given by
(107)
where a and b are positive material constants and h is the invariant defined by (8)1 or, after
use of the incompressibility condition, by (91h. Contours of constant of W(II), when plotted
in the (,\ 1 , ,\ 2 ) plane are the same as contours of constant h, albeit with a different constant.
Such contours are plotted in Figure 6 for several constant values of h. The contours are clearly
convex, and, as we shall see in Section 3.4, this convexity is a consequence of material stability
under dead-loading conditions. If the corresponding contours are plotted, instead, in the plane of
constant (e 1 , e 2 ), where
(108)
are the principal components of the Green strain tensor (9) then this convexity is also evident, as
shown by Holzapfel et al. (2000).
This situation contrasts with that for the (anisotropic) Fung-type strain-energy function (Fung
et al., 1979) defined, in notation adapted to the present context, by
W(Q) = ~ [exp(Q)- 1J, (109)
where p, is a material constant and Q is specified as
(110)
where a, b, care dimensionless material constants and (e 1 , e 2 ) are as given in (108). Contours
of constant W(Q) are also those of constant Q. In the (e 1 ,e 2 ) plane these contours are convex
1 2
Figure 6. Plot of contours of constant !1 in the (.A1, .A2) plane.
provided a > 0, c > 0, ac > b2 , and an example of the plots is shown in Figure 7(a)(necessarily,
e 1 > -1/2 and e 2 > -1/2, corresponding to .\. 1 > 0 and .\.2 > 0). However, it does not in
general follow that the contours of constant Q are convex in (.\. 1 , .\. 2 ) space for the same values
of a, b, c. This is illustrated in Figure 7(b). The lack of convexity implies, as we shall show in a
more general setting in Section 3.3, that a material with the given values of a, b, c may become
unstable under dead loading conditions.
(a) (b)
e2 2
.X2
2
0 1 e1 2 1 .A1
Figure 7. Plot of contours of constant Q in the (e1, e2) plane (a) and the (.A1, .X2) plane (b).
The point that needs to be made here is that special care should be taken in developing forms
of strain-energy function for the modelling of soft tissue (and, of course, for other materials).
Convexity in one strain measure (such as Green strain) does not necessarily guarantee material
88 R. W. Ogden
stability. Equally, lack of convexity may not preclude material stability, but failure of convexity
may have undesirable consequences for the development of numerical schemes for the solution of
(initial-) boundary-value problems. We shall elaborate on these points in the following sections.
3.2 The notion of stability for simple tension

We recall from (54) that for simple tension the Biot stress tis given in terms of the stretch>. by
t = W'(>.). (Ill)
With reference to Figure 1 we see that t is a monotonic increasing function of >. for the two
materials considered there. Thus,
:~ = W"(>.) > o. (112)
This is a mathematical statement of the fact that W is (locally) strictly convex as a function of
>.. If strict inequality is replaced by 2: in (112) then W is (locally) convex. If (112) holds for all
>. > 0 then W is (globally) strictly convex as a function of>., and the curve of W as a function
of>. is bowl shaped upwards and any straight line joining two points of the curve lies above the
curve. The extension of this idea to W(>. 1 ,>. 2 ) will be considered in Section 3.4 but first we
examine how the above considerations change when the stress and strain variables change.
Let e = (>. 2 - 1)/2 and letT = W' (e) be the corresponding component of the second-Piola
Kirchhoff stress (the specialization of (27) 2 ), where W (e) =
W(>.). Then T = >. -lt and
W"(>.) = >. 2 W"(e) + W'(e) = >. 2 ~: + T. (113)
Thus, in tension, for example, the curve ofT as a function of e can have the non-convex shape
shown in Figure 8 without violating the inequality (112). This is another manifestation of the
discussion in Section 3 .1.
0 e
Figure 8. Plot ofT as a function of e.
For homogeneous simple tension under dead load equation (111) may be obtained in an
elementary way by considering the variation with respect to >. of the 'energy function' E(>.)
defined by
E().) = W(A)- t(A- 1), (114)
this being the excess of the stored energy over the work done by the load t in extending the
specimen to a stretch A. The first variation 8E of(ll4) with respect to A is simply
8E(A) = (W'(A)- t)8A, (115)
where 8A is the variation in A. Clearly, E(A) is stationary if and only if ( 111) holds. The second
variation of E(A) is then easily seen to be 82 E = W 11 (A)(8A) 2 , and stability (under dead load)
for this one-dimensional situation therefore requires that (112) holds. In this one-dimensional
situation convexity ofW and (material) stability are equivalent. This is not in general the case in
two or three dimensions, as we shall see in Section 3.4. Incidentally, the term -1 in (114) can be
omitted without affecting the results since it corresponds to the addition of a constant to E(A).
It is worth noting in passing how the above arguments are modified if the loading is not
'dead'. Suppose that the loading is instead of 'follower' type. Specifically, suppose that it cor-
responds to the one-dimensional (tension) specialization of the pressure loading condition (41).
Then we may write t = fA - 1 , where f > 0 is independent of A. The counterpart of ( 114) in this
case is easily shown to be
E(A) = W(A) - J In A, (116)
and the stability inequality (112) is modified to
AW"(A) + W'(A) > o, (117)
which, for tension, is less restrictive than (112).
3.3 Stability in pure homogeneous strain
We next consider an isotropic strain-energy function expressed in terms ofthe principal stretches.
Then, analogously to (114), we consider the total energy
3
E(A1,A2,A3) = W(A1,A2,A3)- :~.::>iAi (118)
i=1
in respect of an unconstrained material. For dead loading, stationarity of (118) with respect to
the (independent) stretches leads to
aw
ti = OAi i E {1, 2, 3}. (119)
ForE to be a (local) minimum in respect of ( 119), i.e. for the second variation of E in the stretch
variations 8A1, 8A2, 8A3 to be positive, we must have
(120)
90 R. W. Ogden
and hence the Hessian matrix (Wij) is positive definite, where a subscript i indicates partial
differentiation with respect to the stretch .\i.
If the stretch variations satisfy the incompressibility constraint
3
L 6.\d Ai = 0, (121)
i=1
obtained by taking the variation ofln(.\ 1 .\2 .\ 3 ) = 0, (120) can be expressed as
(.\iWn- 2.\1.\3W13 + .\~W33)(6.\1/.\1) 2

+ (.\~Wn - 2.\2.\3 W13 + .\~W33)(6.\2/ .\2) 2
+ 2(.\1.\2 W12- .\1.\3 W13- .\2.\3 W23 + .\~W33)(6.\16.\2/ .\1.\2) > o. (122)
Further, by taking .\1, .\2, .\3 to correspond to an isochoric deformation and adopting the notation
defined in (50), (122) may be simplified to give
(123)
where a 3 = .\ 3 W 3 . In fact, it may be shown that (123) is the appropriate modification of (120)
for an incompressible material provided 0'3 is set as 0'3 = .\a w3 - p, where p is the arbitrary
hydrostatic pressure associated with the incompressibility constraint. For a direct derivation of
(123) for an incompressible material we refer to Ogden (1997).
Thus, the inequality (123) is the dead-loading stability requirement for a pure homogeneous
strain subject to the restriction that departure from a state of pure homogeneous strain is not
allowed. Further restrictions are needed if deformation variations that allow the orientation of
the principal axes of strain to change are admitted. This will be discussed in Section 3.5.
For the special case in which a 3 = 0, the inequality (123) reduces simply to the requirement
Hessian matrix is positive definite. (124)
This condition is appropriate, for example, for a thin sheet of material under plane stress and
biaxial tension.
The inequalities (123) and (124) apply not just for isotropic materials but also for anisotropic
materials constructed on the basis of (98) under the restriction to pure homogeneous strain. If
a 3 =f. 0 then (123) puts restrictions on the range of values of a 3 that can be supported in a given
deformed configuration.
The distinction between results for different strain measures can now be brought out very
clearly, and we illustrate this, as in the case of simple_ tension, in respect of the Green strain
measure. Let (TV;j) denote the Hessian matrix of W with respect to the principal Green strains
(e 1, e 2) and let 6e 1, 6e 2 be the variations in (e 1, e2) associated with 6>..1, 6>.. 2. It follows that
2 2
L:wij6>..;6>..j = L:W;j6ei6ei + T¥1(6.\1) 2 + T¥2(6>..2) 2. (125)
i=1 i=1
Thus, we emphasize that local (strict) convexity ofTV(e 1 , e 2) does not in general imply material
stability under dead loading, i.e. local (strict) convexity of W (.\ 1 , .\ 2 ) in the present (restricted)
context since the principal stresses TV1 and TV2 may be negative. In the special case of a thin
sheet under tension, however, the implication does follow. On the other hand, failure of local
(strict) convexity of W (e 1 , e 2 ) does not necessarily mean failure of stability since the stresses
w1 and w2 may be positive.
3.4 Convexity connections

Consider the contour in (,\ 1 , .\ 2 ) space defined by
(126)
where c is a positive constant. On differentiation of ( 126) we obtain
(127)
where (,\~, ,\~) is the unit tangent vector to the contour (126), the prime indicating differentiation
with respect to the arclength parameter.
The curvature of ( 126), denoted "'· is given by the standard formula
(128)
On using both (127) and its derivative with respect to the arclength parameter in ( 128) we obtain,
after some rearrangement,
(129)
This shows the connection between "' and the quadratic form based on the Hessian (Wij).
Note, however, that in (129) the vector (,\~, ,\~) is not arbitrary but is restricted to the tangen-
tial direction to (126). Note also that the sign of"' is dependent on the sense of description of
the contour. Here, this description is taken in the anticlockwise sense so that, for example, the
curvature of a circle of radius a with centre at the origin is + 1/ a.
If the contour (126) is convex then"' > 0 and it may be deduced from (129) that
(130)
since, as we shall show below, the inequality ,\~/W1 > 0 may be adopted independently of (130).
We emphasize that in (130) (,\~,,\~)is restricted as indicated above and hence it does not follow
that the Hessian (Wij) is positive definite. On the other hand, if (Wij) is positive definite then
(130) follows, "' > 0 and the contour is convex. Thus, if the contour is not convex the stability
inequality
2
2:::: Ww5-'i8Aj > o, 8.\i =t o, (131)

i,j=1
obtained from (123), cannot hold. A check on the convexity of the contour (126) therefore pro-
vides a quick test of the suitability of a chosen form of strain-energy function.
92 R. W. Ogden
To illustrate this we consider the neo-Hookean strain-energy function
(132)
where J-1 > 0 is the shear modulus of the material. This is a special case of (69), corresponding
to n = 2. Then,
(133)
from which it follows that
WI 2:0 (~ 0) when .Ai.A2 2: 1 (~ 1),

w2 2: 0 (~ 0) when AIA~ 2: 1 (~ 1). (134)
It is easy to show that (Wij) is positive definite and that, with reference to (127), .A~/WI > 0
and .AUW2 < 0. In Section 3.5.1 we shall conclude that the latter inequalities are valid more
generally, not just for (132). Figure 9 shows a typical contour of constant W(.AI, .A2), as in Figure
6, together with the curves .Ai A2 = 1 and AIA~ = 1, corresponding to WI = 0 and w2 = 0
respectively. Note that these curves cut the contour where the tangent is either horizontal or
vertical and that the contour is symmetric about the line AI = .A 2 .
2 I
--~------~
1 2 3
Figure 9. Contour plot for the neo-Hookean strain energy (132) and the (dashed) curves AiA2 = 1 and
A1A§ = 1 in the (AI, A2) plane.
For comparison with Figure 9 we show, in Figure 10, a corresponding picture for a typical
(convex) anisotropic strain energy based on (1 03). The lack of symmetry should be noted in this
case.
I
I
1 ',L
1-------
\
--r------
1 2
Figure 10. Contour plot for a typical member of the class of anisotropic strain-energy functions (103) and
the (dashed) curves W1 = 0 and W2 = 0 in the ( J\1, J\2) plane.
3.5 Stability under dead loading: the general case
In deriving the inequality (120) we have discounted the possibility that the orientation of the
principal axes of strain can change during. If this is to be allowed for then ( 118) must be replaced
by either
E(U) = W(U)- tr(TU) (135)
or
E(F) = W(F)- tr(SF), (136)
where tr denotes the trace of a second order tensor. The first variations of these with respect to U
and F respectively lead to equations (26h and (18). Note that the difference between these lies
in the fact that (136) allows for rigid rotations through Rand its variation. The conditions for
stability are respectively (at fixed T)
tr {(.CJU)JU} > 0 (137)
for all JU =1- 0 and (at fixed S)

tr { (AJF)JF} > 0 (138)
for all 8F =1- 0, where the fourth-order tensors .C and A are defined by
(139)
The component form of A is given by (44) and that for .C is defined analogously.
It is well known that the inequality (13 8) cannot hold in all configurations, and hence that A
is singular in certain configurations when regarded as a linear mapping on the (nine-dimensional)
94 R. W. Ogden
space of variations JF. We refer to, for example, Ogden (1991, 1997, 2000) for detailed analysis
of the singularities of A and their implications for bifurcation in the dead-load problem.
Modification of these results in the incompressible case requires use of the variations of the
incompressibility condition det F = det U = 1, in the form
(140)
but we do not give details for the general case here.
Specialization to isotropy For the important special case of an isotropic material it is useful to
give explicit expressions for the components of A. The (non-zero) components of A referred to
the principal axes u (i) and v( i) are given by
AiiJJ = WiJ• (141)

Wi+WJ
AiJiJ - AiJJi = Ai + AJ i -1- j, (142)
_ Wi-WJ
AJiJ + AiJJi- A·_ A· i -1- j, Ai -1- AJ, (143)
t J
AiJiJ + AiJJi = Wii - WiJ i -1- j, Ai = AJ, (144)
where Wi = 8Wj8Ai, Wij = 8 2 Wj8Ai8Aj, i,j E {1, 2, 3}, and no summation is implied
by the repetition of indices. For details of the derivation of these components we refer to Ogden
(1997). Equations (141 )-(144) apply for both compressible and incompressible materials subject,
in the latter case, to the constraint (1 0).
In the case of an isotropic material the local stability inequality ( 13 8) can be given explicitly
in terms of the derivatives of the strain-energy function with respect to the stretches. This leads
to
3
tr {(MF)JF} = 2::: wijJAiJA1 + L:(wi- wj)(Ai- Aj) ( nt + ~nn) 2
i,j=l i#j
+ ~ L:(wi + wj)(Ai + Aj)(nm 2 > o, (145)

if.j
where Dt = u(i) · JuUl are the components of the (antisymmetric) tensor ilL that defines the
variation Ju(i) through Ju(i) = nLu(i), and nR = RT JR is also antisymmetric, JR being the
variation in R.
Since JAi, nt, nn are independent, necessary and sufficient conditions for (145) to hold
are therefore
Hessian matrix (Wij) is positive definite, (146)

W-W
t J > 0 i ¥- j, (147)
Ai- AJ
wi + wj > o i -1- i (148)
jointly fori, j E {1, 2, 3}. Note that when .Xi = Aj, i =j:. j (147) reduces to Wii - Wij > 0.
The inequality (146) is just that arising from (120). In respect of (137) only (146) and (147)
are required. Thus, allowing for the change in orientation of the Lagrangian principal axes
imposes the additional inequalities (147) over and above (120). These are obtained by setting
JR = 0 in (145). It is worth noting, however, that it can be shown that (147) follow from (146) if
the region of (.X 1 , .X 2 , .X 3 ) space on which ( 146) holds is convex. The additional inequalities ( 148)
arise solely from geometrical considerations associated with the rigid rotation R.
For the case of plane stress of a thin sheet we may take W3 = 0 and it then follows from
(148) that wl > 0, w2 > 0 and then from (147) that AI > A3 and AI > A3.
For an incompressible material the counterparts of (146)-(148) are
matrix is positive definite, (149)
(150)
(151)
Note thatp occurs in (149)-(151) implicitly through (} 3 and ti, i E {1, 2, 3}.
For the case of plane stress with t 3 = 0 we may deduce that W1 > 0, W2 > 0 and .Xi .X 2 >
1, .X 1 .X~ > 1 as was discussed in Section 3.4 in respect of the neo-Hookean strain energy. The
inequalities (150) on their own are entirely consistent with the factor A.~/W1 in (129) being
positive since, by (127), W1 and .X~ change sign together where the curve .Xi .X 2 = 1 crosses the
contour (126). Indeed, it can be shown that this factor is positive even for non-convex contours,
an example of which is given by Ogden and Roxburgh (1999).
3.6 Elastic moduli in the classical limit

In the classical theory of elasticity, corresponding to the situation in which there is no underlying
deformation or stress, the components of A for an isotropic material can be written compactly
in the form
(152)
where A. and !J are the classical Lame moduli of elasticity and bij is the Kronecker delta. The
values of Wij when .Xi = 1 fori E { 1, 2, 3} are simply Wii = A. + 2tJ, Wij = A., i =f. j. Also,
we take Wi = 0 when A.j = 1 fori, j E {1, 2, 3} so that the configuration Br is a stress free
(natural) configuration.
The counterpart of (152) for an incompressible material is
Aiiii = Aijij = !J, Aiiii = Aiiii = 0 i i- j, (153)
and Wii = Wi = !J, Wii = 0, where !J is the shear modulus in Br. The expressions in (153)
are not uniquely defined because they depend on the point at which .X 1 .X 2 .X 3 is set to unity in the
differentiations in (141)-(144) prior to setting .X 1 = .X 2 = .X3 = 1. In terms of the strain-energy
function W defined in (50) the restrictions required in Br may be written
W(1, 1) = o, Wa(1, 1) = o, l¥12(1, 1) = 2tJ, Waa(l, I)= 4p, (154)

96 R. W. Ogden
where the index a is 1 or 2.

Finally in this section we note that for an anisotropic material more material constants are
needed to characterize the material properties in the classical limit than are given for the isotropic
case. For example, for the strain energy function defined by (98), Wn (1, 1, <p ), w22 (1, 1, <p) and
wl2(1, 1, <.p) are in general independent constants, as can be checked from the example (103).
4 Residual Stresses and Arterial Wall Mechanics
When a length of artery is excised from a body it contracts in length. The resulting configuration
of the specimen is referred to as an unloaded configuration, i.e. it is subject to no axial load nor to
any tractions on its inner and outer surfaces. However, in this configuration there remain stresses
through the artery wall. This is demonstrated by cutting radially a short length of artery in the
form of a ring. The ring springs open to form an open sector (Vaishnav and Vossoughi, 1983;
see, also, Fung, 1993). In general even this open sector is not stress free since the opening angles
of circumferentially separated layers are different (Vossoughi eta!., 1993, and Greenwald eta!.,
1997).
In most analyses, however, the opened-up sector is assumed, for simplicity, to be stress free
in order to facilitate calculation of the (residual) stress required to re-form the intact ring (the un-
loaded configuration) as in, for example, Fung and Liu ( 1989), Delfino eta!. ( 1997) and Holzapfel
et a!. (2000). It is normally assumed that the ring is a circular annulus and that the opened-up
sector is also circular and that the deformation required to re-form the ring depends only on the
radius. It should be noted, however, as we show in Section 4.4, that these kinematic assumptions,
when coupled with the equilibrium equations and boundary conditions, necessarily require the
opened-up sector to be stress free. Any assumptions that are less simple than these would almost
certainly require a purely numerical treatment.
The residual stresses have an influence on the overall behaviour of an artery under extension
and internal pressure and, more significantly, on the stress and strain distributions through the
arterial wall. It has been suggested in the literature that in the physiological state a healthy artery
has an essentially constant circumferential stress in each layer of its wall. This can only be the
case if there is residual stress present (see the discussion in Rachev and Hayashi, 1999, for ex-
ample). It has also been conjectured (Takamizawa and Hayashi, 1987) that the strain distribution
is constant through the wall. Some consequences of the assumptions of uniform circumferential
stress and/or uniform strain have been examined by Ogden and Schulze-Bauer (2000).
The opening angle experiment gives only a very rough estimate of the residual stress, and
a detailed understanding of the mechanical influence of residual stress therefore remains to be
developed. Influences that need to be accounted for are, inter alia, growth, remodelling and
adaptation since these clearly generate residual stresses. Analysis of such effects is at an early
stage of development and much more needs to be done in this area. Recent contributions to these
issues are, for example, Rodriguez eta!. (1994), Rachev (1997, 2000), Rachev eta!. (1998) and
Taber and Humphrey (200 1). See, also, the chapter by Rachev (200 1) in this volume.
Some fundamental aspects of the influence of residual stress on the constitutive law of a
nonlinearly elastic solid have been examined in a series of papers by Hoger and co-workers (see,
for example, Hoger, 1985, 1993, and Johnson and Hoger, 1998). We begin the discussion of
residual stress, in Section 4.1, by examining briefly some of these basic theoretical issues.
4.1 Elastic response in the presence of residual stress

In Section 1.3 we imposed the restrictions (22) on the strain-energy function. Now we suppose
that the reference configuration Br is not stress free, i.e. there is a residual stress in Br and (22)2
no longer holds. Let u(r) denote the residual (Cauchy) stress in Br. In general, this is not obtained
from a strain energy, and we may take the strain-energy function W to be measured from Br and
to satisfy (22)1. Since there is no deformation in Br there is no distinction between the Cauchy
stress and the nominal stress g(r) there. The residual stress must satisfy the equilibrium equation
Div g(r) = 0 in Bn (155)
and, since the boundary is load free, the boundary conditions
g(r)T N = 0 on 8Br. (156)
Since
Div(s(r)®X) = (DivS(r)) ®X+S(r), (157)
it follows from (155), (156) and by application of the divergence theorem that
(158)
An immediate consequence of (158) is that residual stress cannot be uniform- the residual
stress distribution is necessarily inhomogeneous and is therefore geometry dependent. A further
consequence is that the material response of a residually stress body relative to the residually
stressed configuration, and hence the constitutive law, is geometry dependent and inhomoge-
neous.
Residual stress places restrictions on the material symmetry in Br and, in view of the above
remarks, the material symmetry may therefore vary from point to point within the considered
material body. The constitutive laws resulting from these restrictions are, in general, very com-
plicated, and we shall not consider this issue here. We shall, however, examine the simpler issue
of what restrictions are imposed on the residual stress by specific material symmetries.
We restrict attention to symmetry groups that consist of elements of the proper orthogonal
group. Suppose that Q is a rotation tensor belonging to a symmetry group relative to Br. Then,
by combining the stress-deformation relation (18) with the objectivity requirement (24) and the
material symmetry requirement (28) suitably specialized, we obtain
H(QF) = H(F)QT, (159)
for all proper orthogonal Q, and

H(FQ) = QTH(F), (160)
for all members Q of the symmetry group. By setting F =I and g(r) = H(I) and using (159)
and (160), we obtain
g(r)Q = QS(r), (161)
or, equivalently,
(162)
98 R. W. Ogden
where a~k, a~~' a~~ are the components of lT(r).

If there is no dependence on 8 then it follows from ( 166) and the boundary conditions on the
cylindrical surfaces that a~~ = 0 and hence that k 1 and k 2 coincide with the polar coordinate
axes and a ~k = air), a~~ = a ~r). Equations ( 165) and ( 166) then reduce to the single equation
dal(r) (r)
al
(r)
- a2
dR + R = O, (167)
which must be coupled with the boundary conditions
air) =0 onR=A,B. (168)
We shall use equations ( 167) and (168) in Section 4.2 in an analysis of the effect of residual stress
on the response of a circular cylindrical tube under extension and inflation.
4.2 Extension and inflation of a thick-walled tube

We now return to the problem considered in Section 1.5.2 for an isotropic material and Section
2.1.2 for an orthotropic material, but with residual stresses incorporated. The strain energy may
again be written in the form ( 100). Thus, W(A., Az, <p ), with A and Az being the azimuthal and
axial stretches. We emphasize again that W(.A, Az, <p) is not in general symmetric in .A and Az
and that the angle <p may depend on R.
The principal Cauchy stress differences are given (locally) by
(169)
Residual stresses associated with the unloaded configuration may be incorporated through W,
in which case the residual stress differences are given by (169) evaluated for .A = Az = 1 and
subject to a~r) = 0, as discussed above. Alternatively, and this is the approach we adopt here,
the additional stresses required to deform the material from the unloaded configuration may be
accounted for through W via ( 169) with the (in general unknown) residual stresses incorporated
separately. We therefore replace (169) by
(170)
where air), a~r), a~r) = 0 denote the residual principal Cauchy stresses in the unloaded config-
uration where the terms in W vanish. Note that air) and a~r) are independent of the deformation
from the unloaded configuration (i.e. they depend only on R).
For the considered cylindrically symmetric deformation the (radial) equilibrium equations
for the deformed and unloaded configurations are, respectively,
d (r) 1
~ - ( (r) _ (r)) _ O
dR + R al a2 - (171)
in terms of the principal Cauchy stresses. The solution of equation (171 h should satisfy the
boundary conditions
-P on r =a
al = { 0 on r = b,
(172)
where akk, akb, a~b are the components of u(r).

If there is no dependence on 8 then it follows from (166) and the boundary conditions on the
cylindrical surfaces that akb = 0 and hence that k 1 and k 2 coincide with the polar coordinate
axes and ak1 =air), a~b = a~r). Equations (165) and (166) then reduce to the single equation
d a1(r) (r) (r)

a1 - a2
dR + R = O, (167)
which must be coupled with the boundary conditions
air) =0 on R =A, B. (168)
We shall use equations ( 167) and ( 168) in Section 4.2 in an analysis of the effect of residual stress
on the response of a circular cylindrical tube under extension and inflation.
4.2 Extension and inflation of a thick-walled tube

We now return to the problem considered in Section 1.5.2 for an isotropic material and Section
2.1.2 for an orthotropic material, but with residual stresses incorporated. The strain energy may
again be written in the form (100). Thus, W(A, Az, If), with A and Az being the azimuthal and
axial stretches. We emphasize again that W (A, Az, 'P) is not in general symmetric in A and Az
and that the angle 'P may depend on R.
The principal Cauchy stress differences are given (locally) by
(169)
Residual stresses associated with the unloaded configuration may be incorporated through W,
in which case the residual stress differences are given by (169) evaluated for A = Az = 1 and
subject to a~r) = 0, as discussed above. Alternatively, and this is the approach we adopt here,
the additional stresses required to deform the material from the unloaded configuration may be
accounted for through W via (169) with the (in general unknown) residual stresses incorporated
separately. We therefore replace (169) by
- 'W'>.+a2(r) -a1'
a2-a1-A (r)
(170)
where air), a~r), a~r) = 0 denote the residual principal Cauchy stresses in the unloaded config-
uration where the terms in W vanish. Note that air) and a~r) are independent of the deformation
from the unloaded configuration (i.e. they depend only on R).
For the considered cylindrically symmetric deformation the (radial) equilibrium equations
for the deformed and unloaded configurations are, respectively,
1 d (r) 1
da1
- + -(a1 - a2) = 0, ~ + -( (r) _ (r)) _ O (171)
dr r dR R a1 a2 -
in terms of the principal Cauchy stresses. The solution of equation ( 171) 1 should satisfy the
boundary conditions
-P on r =a
a1 = { 0 on r = b,
(172)
100 R. W. Ogden
corresponding to pressure P ( 2:: 0) on the inside of the tube and zero traction on the outside. The
corresponding boundary conditions for the residual stress are (168).
By making use of(58)-(61) together with (170)2 and (171)2, integration of(171h and appli-
!B
cation of the boundary conditions (172) we obtain
P= 1 >-a
Ab
(A 2 A -1)- 1 -dA A- 1
z A + z
awA
a A
R2 d
_
r2
(r)
_!l_dR
dR )
(173)
where, as in (67), the independent variable has been changed from r to A in the first integral,
while in the second integral r 2 is given by (58)1. When there is no residual stress the result (101)
is recovered.
Since, from (60), Ab depends on Aa, equation (173) provides an expression for Pas a function
of Aa when Az is fixed provided that the distribution of residual stress is known. In order to hold
Az fixed an axial load, N say, must be applied to the ends of the tube. This is given by
>. A A
Nf'rrA 2 = (A~Az- 1) { a (A 2 Az -1)- 2 ( 2Az ~7 -A 88:)AdA + PA~

j>.b z
+A;1 iB(2a~r) -a~r) -air))RdRjA2, (174)
and, as for P, this can only be calculated if the residual stress is known. The term in a~r) (= 0)
has been retained so as to maintain the stress difference structure. Note that equation (68) is
recovered when there is no residual stress.
The formulas (173) and (174) are valid for a tube with any number of concentric layers and
for a general strain energy with the specified symmetry. In general, W will be different for each
layer, or, at least, the angle t.p will be different in each layer. The radial stress is continuous across
the boundary between two layers but the circumferential stress is in general discontinuous at such
a boundary.
At this point the residual stress distribution is unknown, and, therefore, to proceed further
we require some means of estimating it. For this purpose some additional information is needed.
One possible approach is to take the opened-up sector of an arterial ring after a radial cut to
correspond to the unstressed configuration (Rachev and Hayashi, 1999, Holzapfel eta!., 2000)
and to investigate the consequences of this assumption. We shall examine some aspects of this in
Section 4.4. One alternative, which we shall consider in Section 4.3, is to investigate the conse-
quences of the assumption that the circumferential stress is uniform at the normal physiological
pressure.
4.3 Uniform circumferential stress

For simplicity of illustration we restrict attention here to a tube with a single layer, but the analysis
(although somewhat more complicated) can easily be carried over to a tube with two or more
layers. If the circumferential stress a 2 = a 20 is assumed to be constant then it follows from the
equilibrium equation ( 171 ) 1 and the boundary conditions ( 172) that
Poao
a2o = b , (175)
o - ao
where the zero subscript indicates evaluation at the normal physiological pressure (Po) and
ro2 = ao2 + "'zo

,-1(R2 - A2) · (176)
Use of equations (168), (170)2, (171 h and (175) then enables the residual radial stress to be
calculated explicitly as
(r)_Poaobo_1_ 1 ((ro-co)(ao+co))-!R'W' ( ' , )dR ( 177 )

CT 1 - og ( )( ) "'O A "'O, AzO, 'P ,
bo - ao 2co ro + co ao - co A R
where c0 = (a6- A-;01 A2 ) 1 12 . The corresponding residual circumferential stress is then obtained
using ( 171 h. This leads to
(178)
Once the residual stresses have been calculated for any given form of W, the pressure P in
a general (cylindrically symmetric) configuration can be calculated from ( 173) and the stresses
from (170) and (171)1. Since CT~r) = 0 the axial load N can be obtained from (174).
By applying the boundary condition (168) at R = B to (177) we obtain
00
P a bo 1 1 ((bo- co)(ao +co)) _ !B AoW (A o, A )dR ( 179)
bo-ao 2co og (b o + co )(ao-co ) - A
A zO, 'P R ·
Since, from (176), b6 = a6

+ A-;01 (B 2 - A2 ), equation (179) provides a connection between
the pressure Po and the internal radius a0 (equivalently, Aoa = a0 I A) for any given value of the
axial stretch Azo and aspect ratio B I A.
A representative plot of the residual stresses is shown in Figure 11 in dimensionless form
with the dimensionless stresses defined by
(J (r)*- CT(r)ll/1. (180)

2 - 2 ,..-3,
where l > 0 is defined by

l = log ( (bo - co) (ao + co) ) (181)
(bo + co)(ao -co)
and f..l 3 is the material constant appearing in the strain-energy function (103), which has been
used in this calculation with n = 12 and f..lr =f..ll/ f..l 3 = 2. The axial stretch Azo has been set at
1.2 and the aspect ratio B I A = 1.2. The general qualitative character of the results in Figure 11
is unchanged by using a range of different values of the material parameters.
We observe that the residual radial stress is quite small and is negative except at the bound-
aries (where it vanishes). The circumferential stress is compressive at the inner boundary and
tensile at the outer boundary and is much larger in magnitude than the radial stress. The results
shown in Figure 11 are very similar to those reported by, for example, Chuong and Fung (1986)
and Takamizawa and Hayashi (1987); see also the discussion in Rachev and Hayashi (1999).
102 R. W. Ogden
0.2
(r )•
(T 1,2
1.1 R/A 1.2
-0.2.
Figure 11. Plot of the dimensionless residual stress distribution for a typical member of the class of
anisotropic strain-energy functions (103) based on equations (177) (radial stress~ dashed curve) and (178)
(circumferential stress ~ continuous curve).
If, in addition to uniform circumferential stress, it is also assumed that the strain distribution
(i.e . .\o) is constant in the physiological state, then ro =.\oR, .\6.\zo = 1, c0 = 0. Further, if the
fibre angle 'P is independent of R then equations (177), (178) and (179) simplify to
(r)/R = ~ (2_ _ 2_) _ log(R/A) (182)

a1 0 B-A A R log(B/A)'
(r) _ B 1 log(R/A)
(183)
a 2 /Po- B- A- log(B/A) log(B/A)'
and
(184)
respectively (Ogden and Schulze-Bauer, 2000). Plots of (182) and (183) illustrated in the latter
paper have the opposite signs to those shown in Figure 11! However, in the work of Rodriguez et
al. ( 1994 ), which takes account of growth, the residual stress has precisely the structure predicted
by (182) and (183) for certain values of their 'growth stretch'.
4.4 The opening angle method

In Figure 12 an arterial ring in three different configurations is depicted. Figure 12 (b) shows the
cross section of an intact artery in the unloaded configuration, while (c) corresponds to an artery
subject to internal pressure P. The deformation from (b) to (c) has been discussed in Section 4.3.
Here, we focus on the deformation from the opened-up configuration, shown in Figure 12 (a), to
the unloaded configuration (b). For reference, we recall that the strain energy associated with the
deformation from (b) to (c) is given by W(.\, Az, <p), where Az (constant) is the axial stretch and
,\ = r / R is the circumferential stretch. The fibre angle in (b) is 'P.
(a) (b) (c)
Figure 12. Opened-up configuration of an arterial ring (a), unloaded intact ring (b) and deformed configu-
ration under pressure P (c).
We assume that the sector in (a) is circular and has an opening angle a, as indicated in the
figure. Note that different definitions of opening angle are sometimes used in the literature. It is
convenient to use the notation
k = 27r/(27r- a), 1:::; k < oo, (185)
as a measure of the opening angle. In the deformation from (a) to (b) we assume that there is
a uniform stretch Azo induced in the axial direction. The radial part of the deformation is then
given by
(186)
where Ro is the radial coordinate in (a) and Ao is the inner radius.
The associated circumferential stretch, denoted A0 , is
(187)
and we denote by r.p 0 the fibre angle in (a). (Note that the notations A0 and Azo differ from the
Ao and Azo used in Section 4.3.) By applying the formula (85) to the deformation (a) to (b) we
deduce that
tan'P = AoX;-01 tan 'Po· (188)
Next, we assume that the deformation from (a) to (b) is an elastic deformation and described
by the strain energy W0 (A 0 , Az 0 , 'Po), where the subscript o is attached toW since, in general, the
material response relative to (a) will be different from that relative to (b) even after accounting for
the change in fibre angle because, in general, the deformation induces anisotropy in the response
relative to (b).
In most analyses it is assumed that the configuration (a) is stress free. We now show that
this assumption is valid since the choice of geometry necessarily leads to (a) being stress free.
Suppose that (a) is not stress free. The geometry ensures that the principal axes of strain are
radial and circumferential. Since the deformation is independent of the polar coordinate angle,
e
denoted 0' it follows that the principal axes of stress coincide with those of strain and that the
only equilibrium equation not satisfied trivially in (a) is the radial equation
(r)
d (lol 1
( (r) (r)) _
dRo + Ro (lol - (lo2 - 0, (189)
104 R. W. Ogden
where ai~) and ai;) are, respectively, the radial and circumferential (residual) principal stresses
in (a). Since the load must vanish pointwise on the (flat) ends of the opened-up ring we must have
a~;) = 0 on those ends (on which 8 0 is constant). It follows from (189) that d(R 0 a~~))/dR 0 = 0
on the ends, and hence for all8 0 • Integration of this and application of the zero traction condition
=
ai~) = 0 on Ro = A 0 shows that ai~) 0 and hence, by (189), a~;) 0. =
This result applies for one layer or for two or more concentric layers, so, in particular, for
the case of two layers, the interface must therefore form a perfect geometrical match in the
configuration (a). In practice this is unlikely, and indeed, as has been shown in experiments in
Professor Holzapfel's laboratory, this is certainly not the case. The length of the outer boundary
of the media is not in general the same as the length of the inner boundary of the adventitia
in the opened-up configuration. Moreover, the curvatures of these boundaries are not in general
the same. For the media and adventitia to fit together in the opened-up configuration there will
necessarily be residual stresses in that configuration. In view of the above analysis such a con-
figuration cannot be described by the geometry discussed above and the deformation from (a)
to (b) must depend on 8 0 , and possibly also on the axial coordinate. The analysis associated
with this more general geometry is, of course, more complicated than described above and will
undoubtedly require numerical treatment. In particular, the plane strain assumption is unlikely to
be a good approximation to the real situation for a short length of artery. Work is still in progress
on this. Here, our analysis is based on (186).
The residual stress distribution in (b) is governed by equation ( 171 )2, which, on integration,
gives
(r) _ {R( (r) (r)) dR
al -}A a2 - al R' (190)
but now the integrand in (190) is given by

- \ WA (' \ )
a2(r) -al(r) -/\o OA Ao,Azo,i{)o•
0
(191)
Thus, in principle, the residual stress can be calculated. However, this requires some additional
information.
First, we note that if Eo denotes the outer radius in (a) then the geometrical quantities in (a)
and (b) are related by
E2 = A2 + k-1.\-l(E2
zo 0
_ A2)
0 • (192)
Secondly, by applying the boundary condition air) = 0 on R = E to (190) we obtain

{B dR
Ao Wo>.JAo, Azo, i{)o) R = 0,
A
} A (193)
or, equivalently, by changing the integration variable from R to A0 using (186) and (187),
r>-oa Wo>.a (Ao, Azo, i{)o) d,\ = 0

(194)
}Aob A~Azo- k 0
'
where Aoa and Aob are the values of A0 on the boundaries Ro = Ao and Ro = Eo respectively.
Since our objective is to calculate the residual stress distribution, we suppose that k, A 0 , Eo
and Azo are known. Equations (192) and (194) are then two equations from which to determine
A and B, the latter equation depending on the material properties through Wand 'Po· Note that
A, B, A 0 , B 0 occur in (194) only through the limits. Once A and Bare determined the residual
stresses can be calculated from (190) and (191). In this way the opening angle can be related to
the residual stress in the unloaded configuration and hence, through the analysis in Section 4.3,
the opening angle required to give uniform circumferential stress in the normal physiological
state can be determined.
In the above considerations we have not made use of the equation
(r) - 0"1(r) --
0"3
' w'OAzo ('Ao, Azo,
Azo ' 'Po )· (195)
This is important to note since the zero axial load condition a ~r) = 0 in (b) is not in general
compatible with the assumed geometrical transformation from (a) to (b). Thus, (195) must be
regarded as giving the stress distribution a ~r) needed to maintain the cylindrical geometry in (b),
in particular uniform Azo· As is done in some treatments, this problem can be circumvented by
setting to zero the total axial load
(196)
so as to determine the value of Azo· Alternatively, Azo can be prescribed and a~r) calculated from
(195) once air) has been determined by the procedure outlined above.
(a) (b)
0.01
0.02
1.2
-0.02
-0.01
Figure 13. Plot of the residual stresses based on equations (190)-(194) in dimensionless form: (a) radial
stress (dashed curve), circumferential stress (continuous curve); k = 1.5: (b) comparison of the residual
circumferential stresses fork = 1.5, 1.6, 1. 7.
Some results based on the latter approach are shown in Figure 13 with Azo set at 1. In Fig-
ure 13 (a) dimensionless radial and circumferential residual stresses are plotted against the di-
mensionless radius Ro/Ao with B 0 /Ao set to 1.2. An opening angle of 21r /3, corresponding to
k = 1.5, has been selected. The plots are based on use of the energy function (103), as for the
results shown in Figure 11, and the general character of the results is the same as in Figure 11.
106 R. W. Ogden
Here, however, the non-dimensionalization used is different and is based on division by the con-
stant IJL3 1. Figure 13 (b) shows a comparison of the circumferential stresses for three different
opening angles, corresponding to k = 1.5, 1.6, 1. 7. Two features should be noted. First, the stress
is compressive on the inner boundary and tensile on the outer boundary; second, the maximum
magnitudes of the stress increase with the value of k. The point at which the stress vanishes is
slightly different for the three curves although this is not apparent on the scale used here. The
radial stress likewise increases with k but remains very small compared with the circumferential
stress and hence the corresponding comparison is not shown.
If it is not assumed that Azo is uniform then the problem becomes more difficult because
the deformation from (a) to (b) then necessarily involves shearing through the wall thickness. A
more general analysis of the opening angle method in which this is accounted for in an axially
symmetric setting is given by Ogden (2002).
References
Beatty, M.F. (1987). Topics in finite elasticity: hyperelasticity of rubber, elastomers and biological tissues-
with examples. Appl. Mech. Rev. 40: 1699-1734.
Chuong, C.J. and Fung, Y.C. (1986). On residual stress in arteries. J Biomech. Eng. 108:189-192.
Coleman, B.D. and Noll, W. (1964). Material symmetry and thermostatic inequalities in finite elastic defor-
mations. Arch. Rational Mech. Anal. 15:87-111.
Delfino, A., Stergiopulos, N., Moore, J.E. and Meister, J.-J. (1997). Residual strain effects on the stress field
in a thick wall finite element model of the human carotid bifurcation. J Biomech. 30:777-786.
Fu, Y.B. and Ogden, R.W. (2001). Nonlinear Elasticity: Theory and Applications. Cambridge University
Press.
Fung, Y.C. (1993). Biomechanics: Mechanical Properties of Living Tissue, 2nd Edition. New York:
Springer.
Fung, Y.C., Fronek, K. and Patitucci, P. (1979). Pseudoelasticity of arteries and the choice of its mathemat-
ical expression. Am. J Physiol. 237:H620-H631.
Fung, Y.C. and Liu, S.Q. (1989). Change of residual strains in arteries due to hypertrophy caused by aortic
constriction. Circ. Res. 65:1340--1349.
Greenwald, S.E., Moore, J.E., Rachev, A., Kane, T.P.C. and Meister, J.-J. (1997). Experimental determina-
tion of the distribution of residual strains in the artery wall. J Biomech. Engr. 119:438--444.
Haughton, D.M. and Ogden, R.W. (1979a). Bifurcation of inflated circular cylinders of elastic material
under axialloading-I. Membrane theory for thin-walled tubes. J Mech. Phys. Solids 27: 179-212.
Haughton, D.M. and Ogden, R.W. (1979b). Bifurcation of inflated circular cylinders of elastic material
under axialloading-II. Exact theory for thick-walled tubes. J Mech. Phys. Solids 27:489-512.
Hoger, A. ( 1985). On the residual stress possible in an elastic body with material symmetry. Arch. Rational
Mech. Anal. 88:271-290.
Hoger, A. (1993). The constitutive equation for finite deformations of a transversely isotropic hyperelastic
material with initial stress. J Elasticity 33:107-118.
Holzapfel, G.A. (2000). Nonlinear Solid Mechanics. Chichester: Wiley.
Holzapfel, G.A. (2001a). Biomechanics of soft tissue. In Lemaitre, J., ed., Handbook of Materials Behavior
Models. Boston: Academic Press. 1049-1063.
Holzapfel, G.A. (2001b). Structural and numerical models for the (visco) elastic response of arterial walls
with residual stresses. In Holzapfel, G.A. and Ogden, R.W., eds., Biomechanics of Soft Tissue. CISM
Courses and Lectures Series. Wien: Springer-Verlag.
Holzapfel, G .A. and Gasser, C. T. (200 1). A viscoelastic model for fiber-reinforced materials at finite strains:
continuum basis, computational aspects and applications. Comput. Meth. Appl. Mech. Engr. 190:4379-
4403.
Holzapfel, G.A., Gasser, C.T. and Ogden, R.W. (2000). A new constitutive framework for arterial wall
mechanics and a comparative study of material models. J. Elasticity 61:1-48.
Humphrey, J.D. (1995). Mechanics of the arterial wall: review and directions. Critical Reviews in Biomed.
Engr. 23:1-162.
Humphrey, J.D. (1999). An evaluation ofpseudoelastic descriptors used in arterial mechanics. J. Biomech.
Engr. 121:259-262.
Johnson, B.E. and Hoger, A. (1998). The use of strain energy to quantify the effect of residual stress on
mechanical behavior. Math. Mech. Solids 3:447-470.
Nichols, W.W. and O'Rourke, M.F. (1998). McDonald's Blood Flow in Arteries, 4th edition, chapter 4, pp.
73-97. London: Arnold.
Ogden, R.W. (1972). Large deformation isotropic elasticity: on the correlation of theory and experiment for
incompressible rubberlike solids. Proc. R. Soc. Lond. A 326:565-584.
Ogden, R.W. (1982). Elastic deformations of rubberlike solids. In Hopkins, H. G. and Sewell, M.J., eds.,
Mechanics of Solids, the Rodney Hill 60th Anniversary Volume. Oxford: Pergamon Press. 499-537.
Ogden, R.W. (1991). Nonlinear elasticity: incremental equations and bifurcation phenomena. In Ames,
W.F. and Rogers, C., eds., Nonlinear Equations in the Applied Sciences. New York: Academic Press.
437-468.
Ogden, R.W. (1997). Non-linear Elastic Deformations. New York: Dover Publications.
Ogden, R.W. (2000). Elastic and pseudo-elastic instability and bifurcation. In Petryk, H., ed., Material
Instabilities in Elastic and Plastic Solids. Wien: Springer-Verlag. 209-259. CISM Courses and Lectures
No. 414.
Ogden, R.W. (2001). Elements of the theory of finite elasticity. In Fu, Y.B. and Ogden, R.W., eds., Nonlinear
Elasticity: Theory and Applications. Cambridge University Press. 1-57.
Ogden, R. W. (2002). An analysis of the opening angle method for the determination of residual stress in
arteries, submitted.
Ogden, R.W. and Roxburgh, D.G. (1999). A pseudo-elastic model for the Mullins effect in filled rubber.
Proc. R. Soc. Lond. A 455:2861-2877.
Ogden, R.W. and Schulze-Bauer, C.A.J. (2000). Phenomenological and structural aspects of the mechanical
response of arteries. In Casey, J. and Bao, G., eds., Mechanics in Biology. New York: ASME. AMD-Vol.
242/BED-Vol. 46:125-140.
Rachev, A. (1997). Theoretical study of the effect of stress-dependent remodeling on arterial geometry
under hypertensive conditions. J. Biomech. 30:819-827.
Rachev, A. (2000). A model of arterial adaptation to alterations in blood flow. J. Elasticity 61:83-111.
Rachev, A. (2001). Remodeling of arteries in response to changes in their mechanical environment. In
Holzapfel, G.A. and Ogden, R.W., eds., Biomechanics of Soft Tissue. CISM Courses and Lectures
Series. Wien: Springer-Verlag.
Rachev, A. and Hayashi, K. (1999). Theoretical study of the effects of vascular smooth muscle contraction
on strain and stress distributions in arteries. Ann. Biomed. Engr. 27:459-468.
Rachev, A., Stergiopulos, N. and Meister, J.-J. (1998). A model for geometric and mechanical adaptation of
arteries to sustained hypertension. J. Biomech. Engr 120:9-17.
Rodriguez, E., Hoger, A. and McCulloch, A.D. (1994). Stress-dependent finite growth in soft elastic tissues.
J. Biomech. 27:455-467.
Schulze-Hauer, C.A.J., Morth, C. and Holzapfel, G.A. (2001). Passive quasistatic multiaxial mechanical
response of aged human iliac arteries, submitted.
Spencer, A.J.M. (1972). Deformations of Fibre-reinforced Materials. Oxford University Press.
108 R. W. Ogden
Spencer, A.J.M. (1984). Constitutive theory for strongly anisotropic solids. In Spencer, A.J.M., ed., Contin-
uum Theory of the Mechanics of Fibre-reinforced Composites. Wien: Springer-Verlag. CISM Courses
and Lectures No. 282:1-32.
Taber, L.A. and Humphrey, J.D. (2001). Stress-modulated growth, residual stress, and vascular heterogene-
ity. J Biomech. Engr. 123:528-535.
Takamizawa, K. and Hayashi, K. (1987). Strain energy density function and uniform strain hypothesis for
arterial mechanics. J Biomech. 20:7-17.
Vaishnav, R.N. and Vossoughi, J. (1983). Estimation of residual strains in aortic segments. In Hall, C.W.,
ed., Biomechanical Engineering II: Recent Developments. New York: Pergamon Press. 330--333.
Vossoughi, J., Hedjazi, Z. and Boriss, F.S. (1993). Intimal residual stress and strain in large arteries. In
Proceedings of the ASME Bioengineering Conference. New York: ASME. BED-Vol. 24:434-437.
Structural and Numerical Models for the (Visco )elastic
Response of Arterial Walls with Residual Stresses
Gerhard A. Holzapfel
Graz University of Technology, Institute for Structural Analysis- Computational Biomechanics

Schiesstattgasse 14-B, A-8010 Graz, Austria
E-mail: gh@biomech.tu-graz.ac.at, Home Page: http://www.cis.tu-graz.ac.at/biomech
Abstract. In this chapter we focus attention on the description of structural and numerical
models for the elastic and viscoelastic response of arterial walls with residual stresses. We
start by reviewing briefly the arterial histology and describing the mechanical characteris-
tics of arterial components. We also present a fully automatic technique for identifying the
orientations of cellular nuclei.
Particular attention is concentrated on multi-layer models for predicting reliably the passive
elastic and viscoelastic three-dimensional stress and deformation states of arterial walls un-
der various loading conditions. All the models proposed are well suited for FE implementa-
tion. Each arterial layer is treated as a fiber-reinforced material with the fibers corresponding
to the collagenous constituent of the material and symmetrically disposed with respect to
the cylinder axis. The resulting constitutive law is orthotropic in each layer. A specific form
of the law, which requires only three material parameters for each layer, is used to study the
response of an artery under combined axial extension, inflation and torsion. The character-
istic and very important residual stress in an artery in vitro is accounted for by assuming that
the unstressed configuration of the material corresponds to an open sector of a tube. The vis-
coelastic model admits hysteresis loops that are known to be relatively insensitive to strain
rate, an essential mechanical feature of muscular arteries. The concept of internal variables
is introduced in order to replicate the characteristic dissipative mechanism. We summarize
the equations that provide the general continuum description of the deformation and the hy-
perelastic stress response of arterial walls, which are assumed to behave isochorically. One
particular simple mixed FE method is discussed in detail (leading to the Ql/ PO-element).
This approach circumvents numerical difficulties that arise from the overstiffening of the
system associated with the analysis of isochoric deformations. Stiffness matrices and some
insights into solution methods for nonlinear dynamic problems are provided.
Three numerical examples are included to show the performance of the structural arterial
models and to document FE results that are in good qualitative agreement with experimental
data. The first example is concerned with a FE analysis of the mechanical behavior of an
artery during clamping. The remaining two examples are concerned with investigation of
the characteristic viscoelastic behavior of a healthy young artery under static and dynamic
boundary loadings.
In the last section a layer-specific FE-model of balloon angioplasty is described. The model
makes use of an in vitro MRl of a human stenotic post-mortem artery and mechanical
tests of the corresponding vascular tissues under supra-physiological loadings. The three-
dimensional FE realization considers the balloon-artery interaction and accounts for vessel-
specific axial in situ pre-stretches. The proposed approach provides a tool that has the po-
tential to improve procedural protocols and the design of interventional instruments on a
lesion-specific basis, and to determine post-angioplasty mechanical environments, which
may be correlated with restenosis responses.
110 G.A. Holzapfel
1 Arterial Histology
This section aims to review briefly the histology of arteries and to describe the mechanical char-
acteristics of arterial components that provide the main elastic and viscoelastic contributions to
the arterial deformation process. In addition, we review a fully automatic technique for identi-
fying the orientations of cellular nuclei, these coinciding with the preferred orientations in the
tissue. The distribution of preferred (fiber) orientations provide important histological informa-
tion for the structural models documented in Sections 2 and 3.
1.1 Histology and Mechanical Characteristics of Arterial Components
Arteries are blood vessels with a wide variety of diameters. They are roughly subdivided into two
types: elastic and muscular. Elastic arteries such as the aorta and the carotid are located close
to the heart (proximal arteries), have relatively large diameters and may be regarded as elastic
structures. Muscular arteries (distal arteries) such as femoral, coeliac and cerebral arteries are
smaller, located at the periphery (more distal) and may be regarded as viscoelastic structures.
Smaller arteries typically display more pronounced viscoelastic behavior than arteries with large
diameters; see, for example, Tanaka and Fung (1974). It is generally assumed that the content of
smooth muscle cells present in an artery is responsible for its viscosity. For example, according
to the study by Learoyd and Taylor (1966), high viscosity is found in the human femoral artery,
and this is attributed to its very large muscle content.
Arterial walls are composed of three distinct concentric layers, the intima (tunica intima), the
media (tunica media) and the adventitia (tunica externa). The mechanical properties of arterial
walls are strongly influenced by the concentration and structural arrangement of constituents
such as collagen and elastin, the hydrated matrix of proteoglycans, and the topographical site
and respective function in the organism (see, for example, the early work by Roy (1880-82)).
The tensile responses of circumferential and longitudinal specimens vary along the aortic
tree (Bergel (1961b), Bergel (1961a), Learoyd and Taylor (1966) and Tanaka and Fung (1974)
among many others). Non-diseased arterial walls (with straight vessel axes) behave cylindrically
orthotropically, which is well-established in the literature; see, for example, the early work by
Patel and Fry (1969). For mechanical properties and constitutive equations of arterial walls, see
the reviews by, for example, Hayashi (1993), Humphrey (1995), the data book edited by Abe
et al. (1996), Section 2, and Holzapfel et al. (2000a) and Humphrey (2002). In the following
we discuss the constituents of arterial walls from the mechanical perspective. Figure 1 shows a
model of a healthy elastic artery.
Intima. The intima is the innermost layer of the artery. It consists of a single layer of endothe-
lial cells lining the arterial wall and resting on a thin basal membrane (basal lamina). There is also
a subendothelial layer whose thickness varies with topography, age and disease. In healthy young
muscular arteries, however, the subendothelial layer is almost non-existent. In healthy young in-
dividuals the intima is very thin and makes an insignificant contribution to the solid mechanical
properties of the arterial wall.
However, the mechanical contribution of the intima may become significant for aged arteries
(arteriosclerosis) (the intima becomes thicker and stiffer); see for example, Schulze-Bauer et al.
(200 1). In addition, it is important to note that pathological changes of the intimal components
Structural and Numerical Models for the (Visco)elastic Response of Arterial Walls... 111
Composite reinforced by
collagen fibers arranged
in helical structures
Helically arranged fiber-

reinforced medial layers
Bundles of collagen fibrils

External elastic lamina
Elastic lamina
Elastic fibril
Collagen fibrils
Smooth muscle cell
Internal elastic lamina
Endothelial cell
Figure 1. Diagrammatic model of the major components of a healthy elastic artery composed of three
layers: intima (I), media (M), adventitia (A). I is the innermost layer consisting of a single layer
of endothelial cells that rests on a thin basal membrane and a subendothelial layer whose thickness
varies with topography, age and disease. M is composed of smooth muscle cells, a network of
elastic and collagen fibrils and elastic laminae which separate M into a number of fiber-reinforced
layers. The primary constituents of A are thick bundles of collagen fibrils arranged in helical
structures; A is the outermost layer surrounded by loose connective tissue.
(atherosclerosis) are associated with significant alterations in the mechanical properties of arte-
rial walls, differing significantly from those of healthy arteries (Learoyd and Taylor (1966) and
Langewouters eta!. (1984)).
Media. The media is the middle layer of the artery and consists of a complex three-dimensio-
nal network of smooth muscle cells, elastin and collagen fibrils. According to Rhodin ( 1980)
the fenestrated elastic laminae separate the media into a varying number of well-defined con-
centrically fiber-reinforced medial layers. The number of elastic laminae decreases toward the
periphery (as the size of the vessels decreases) so that elastic laminae are hardly present in mus-
cular arteries.
The media is separated from the intima and adventitia by the so-called internal elastic lamina
and external elastic lamina (absent in cerebral blood vessels), respectively. In muscular arteries
these laminae appear as prominent structures, whereas in elastic arteries they are hardly dis-
112 G.A. Holzapfel
tinguishable from the regular elastic laminae. The close interconnection between the elastic and
collagen fibrils, elastic laminae, and smooth muscle cells together constitute a continuous fibrous
helix (Faserschraube), see Schultze-Jena (1939) and Staubesand (1959). The helix has a small
pitch so that the smooth muscle cells in the media are almost circumferentially oriented. This
structured arrangement gives the media high strength, resilience and the ability to resist loads in
both the longitudinal and circumferential directions. From the mechanical perspective, the media
is the most significant layer in a young healthy artery.
Due to the high content of smooth muscle cells, it is the media that is believed to be mainly
responsible for the viscoelastic behavior of an arterial segment.
Adventitia. The adventitia is the outermost layer of the artery and consists mainly of fibro-
blasts and fibrocytes (cells that produce collagen and elastin), histological ground substance and
thick bundles of collagen fibrils forming a fibrous tissue. The adventitia is surrounded contin-
uously by loose connective tissue. The thickness of the adventitia depends strongly on the type
(elastic or muscular) and the physiological function of the blood vessel and its topographical site.
For example, in cerebral blood vessels there is virtually no adventitia.
The wavy collagen fibrils are arranged in helical structures and serve to reinforce the wall.
They contribute significantly to the stability and strength of the arterial wall. The adventitia is
much less stiff in the load-free configuration and at low pressures than the media. However, at
higher levels of pressure the collagen fibers reach their straightened lengths and the adventi-
tia changes to a stiff 'jacket-like' tube which prevents the artery from overstretch and rupture
(Schulze-Bauer eta!. (2002)).
Collagen. Collagen is a protein which is very important for vertebrate physiology. It is a

macromolecule with length of about 280 nm. The rod-like shape of the collagen molecule comes
from three polypeptide chains which are composed in a right-handed triple-helical conformation.
Most of the collagen molecule consists of three amino acids; glycine (33%), which enhances
the stability of the molecule, proline (15%) and hydroxyproline (15%). Collagen molecules are
linked to each other by covalent bonds building collagen fibrils. Depending on the primary func-
tion and the requirement of strength of the tissue the diameter of collagen fibrils varies (the order
of magnitude is 1.5 nm; see Nimni and Harkness (1988)). Collagen appears as concentrically
arranged fibers in the structure of blood vessels. More than 12 types of collagen have been iden-
tified (for comprehensive data see Nimni (1988)). The most common collagen is type I, which
can be isolated from any tissue. It is the major constituent in blood vessels.
The intramolecular crosslinks of collagen gives the tissues the strength which varies with
age, pathology, etc. The function and integrity of organs are maintained by the tension in col-
lagen fibers. They shrink upon heating due to breakdown of the crystalline structure (at 65 a C,
for example, mammalian collagen shrinks to about one-third of its initial length, Fung (1993),
p. 263). Collagen fibers represent the main load carrying elements of arterial walls that render
the material properties anisotropic.
Elastin. Elastin is a protein which is present as thin strands in soft tissues. The ratio of elastin
to collagen in the aorta decreases away from the heart (Wolinsky and Glagov (1967)). The long
flexible elastin molecules build up a three-dimensional (rubber-like) network, which may be
stretched to about 2.5 of the initial length of the unloaded configuration. In contrast to collagen
fibers, this network does not exhibit a pronounced hierarchical organization. As for collagen, 33%
of the total amino acids of elastin consists of glycine. However, the proline and hydroxyproline
contents are much lower than in collagen molecules. Elastin behaves like a rubber band and can
sustain extremely large strains without rupturing (it fractures at low stresses). It is essentially a
linear elastic material (tested for the ligamentum nuchae of cattle- Fung and Sobin (1981 )) and
displays very small relaxation effects (they are larger for collagen).
Further discussion of arterial histology is contained in Chapter 1 by HUMPHREY and Mc-
CULLOCH in this volume. For a more detailed account of the different mechanical characteristics,
the structure (distribution and orientation) of the interrelated arterial components, the morpho-
logical structure and the overall functioning of the blood vessel the reader is referred to, for
example, Rhodin (1980), Silver eta!. (1989), Humphrey (1995), Holzapfel eta!. (2000a), Sec-
tion 2, Holzapfel (200 1) and Humphrey (2002).
1.2 Identification of Preferred Orientations in Arterial Layers

The quantitative knowledge of preferred orientations in arterial layers enhances the understand-
ing of the general mechanical characteristics of arterial walls significantly. It is important to note
that realistic structural models rely strongly on this knowledge. Collagen fibers are those compo-
nents of arterial walls that render the material properties anisotropic. To describe the anisotropic
feature, appropriate geometrical data (fiber angles) are required. They serve as an essential set
of input data for numerical models. In this section we review an automatic technique for ob-
taining information about preferred orientations in isolated arterial tissues, and, additionally, the
concentration of nuclei recently proposed by Holzapfel eta!. (2002a).
Automatic technique for identifying preferred orientations in arterial layers. High direc-
tional correlations between the long axes of cellular shapes and nuclei of, for example, a stained
patch of a media may be studied under a microscope. Since a stained patch reveals the orienta-
tions of smooth muscle cells, in particular of the associated nuclei (see Figure 2), it is possible
to identify the statistical distributions of nuclei in the patch plane. In particular, we detect nuclei
on an algorithmic basis by scanning each line of the histological section (image) consecutively.
All connected pixels that make up a nucleus are copied to a stack. The stack is then transferred
to a program, which determines the orientation of the shape of the nucleus from its associated
second-order moment. To continue, the nucleus so detected is then deleted from the original im-
age, and the procedure starts again at the location where the first pixel of the preceding nucleus
was detected.
The above mentioned program starts with the computation of the centroid Xc of a nucleus
according to Xc = 2:~ 1 x;/N, where N refers to the total number of pixels making up the
nucleus. A representative shape of a nucleus is illustrated in Figure 2. The position vector of a
typical pixel i relative to a fixed origin is denoted by xi and the components of the vectors Xc and
Xi are Xc, Yc and Xi, Yi, respectively.
Knowing the centroid of the nucleus we may compute the second-order moment, G say, i.e.
N
G = I:[(ri · ri)I- ri ® ri], (1)
i=l
where we introduced the definition ri = Xi - Xc of the position vectors ri, and I denotes the
second-order unit tensor which, in index notation, has the form (I) ij = 8ij with 8ij being the
114 G.A. Holzapfel
Close up view
Histological section of a nucleus
Human aortic media
D
Figure 2. Histological section of a stained patch of a human aortic media. The circumferential orientation
of the patch was aligned with the horizontal axis. The algorithm detected 179 nuclei within the
image representing a histological size of0.5 x 0.5mm 2 (pmin = 15, Tmin = 2).
(a) (b)
35
30
25
20
15
10 5
-60° -40° -20° oo 20° 40°

Orientation of nuclei
Figure 3. Histograms show statistical distributions of the orientation of nuclei in a human aortic media (a),
and of collagen fibers in a human aortic adventita (b). Statistical analyses led to mean angles
'PM = ±8.4° in the media and 'P A = ±41.9° in the adventitia.
Kronecker delta. To write eq. ( 1) in a more convenient matrix representation, which is useful for
computational purposes, we have
(2)
If we imagine that every pixel, for example, possesses mass, then G may be seen as the inertia
tensor relative to a fixed origin. It is a symmetric second-order tensor with components forming
the entries of the inertia matrix, denoted by G and expressed through eq. (2). The diagonal
components (xi- Xc) 2 , (Yi- Yc) 2 and the off-diagonal components -(xi- Xc)(Yi- Yc) ofG
are associated with the moments of inertia and the product of inertia, respectively. Considering
the eigenvalue problem for the matrix G we may finally find the eigenvalues (which are the
principal moments of inertia) and the associated eigenvectors which define the corresponding
principal axes. Hence, the orientation of the nucleus is clearly determined by its principal axes,
relative to which the product of inertia is zero. To this end it is immaterial whether only perimeter-
pixels or all the pixels making up the nucleus are considered. Subsequently, the (mean) angle
that occurs between the collagen fibers and the circumferential direction of the respective arterial
layer (media, adventitia) is denoted by cp.
Finally, note that special attention is focused on the type of data stored in a stack. In order
to avoid artifacts such as scratches, which may occur in a histological section, or nuclei that are
represented by almost circular shapes, we require (i) a minimum number of pixels Pmin that make
up a nucleus and (ii) a minimum ratio rmin of larger to smaller eigenvalues (principal moments
of inertia). If both criteria are satisfied the nucleus is accepted and its orientation is considered as
relevant data.
Example. A histological section of a stained patch of a human aortic media with size 0.5 x
0.5mm 2 is considered. The image, as illustrated in Figure 2, was digitized and prepared for
scanning by the software tool developed. The horizontal direction of the patch was aligned with
the circumferential direction of the vessel. The criteria which ensure that the orientation of a
nucleus is considered as relevant data were set to the values Pmin = 15 and rmin = 2.
The algorithm detected 179 nuclei within the image and determined the orientations of all
nuclei. A typical statistical distribution of the orientation of nuclei is illustrated in Figure 3(a)
in the form of a histogram. The two (light grey) peaks in the figure fill the same area as the
histogram. Statistical analyses led to mean angles 'PM of the orientations of nuclei (collagen
fibers) in the media of ±8.4°. The result that the smooth muscle cell orientations in the media
are almost circumferentially oriented is in agreement with general histological data of a human
aortic media (see, for example, Rhodin (1980)).
Figure 3(b) shows the representative results of a statistical analysis performed for a human
aortic adventitia, which is characterized by predominant orientations of collagen fibers. The
mean angles between the collagen fibers and the circumferential direction in the adventitia are,
for this sample, determined as cpA = ±41.9°. The automatic technique for identifying fiber di-
rections in the adventitia is very similar to that described above. Alternatively, for low numbers
of nuclei the intra-spatial voids between collagen fiber bundles may be used as indicators for
preferred orientations of the tissue.
116 G.A. Holzapfel
2 A Structural Model for the Elastic Behavior of Arterial Walls
In this section we propose a multi-layer model for the passive (unstimulated) behavior of healthy
young arterial walls in the elastic deformation domain. The structural model is formulated within
the framework of nonlinear continuum mechanics and is well-suited for a finite element imple-
mentation.
We summarize the equations that provide the general continuum description of the defor-
mation and the hyperelastic stress response of the material. From experimental observations of
arterial walls it is known that their deformation response is almost isochoric in nature (the bulk
modulus considerably exceeds the shear modulus). For this physical reason it seems to be most
beneficial to choose a decoupled representation of the free-energy function. The decoupled rep-
resentation turns out to be advantageous in avoiding numerical complications within the finite
element analysis of incompressible materials (see Section 4). As a basis for reporting the perfor-
mance of the constitutive model we consider the mechanical response of a thick-walled circular
cylindrical tube under various boundary loads. We specify the stretches of the deformation and
the amount of shear to be used, and discuss the equilibrium equation which arises in the consid-
ered problem. We also give expressions for the torsional couple and the reduced axial force acting
on the tube, these being crucial for the numerical study of the constitutive model performed on a
carotid artery. The numerical example in Section 5.1 of this chapter demonstrates one (clinical)
application of the proposed multi-layer model. It documents a large scale finite element analysis
of the elastic behavior of a healthy and young arterial segment during clamping.
2.1 Hyperelasticity
In the following we describe the (finite) deformation and the stress response of hyperelastic
materials.
Description of the deformation. Let Do be a region of a continuum body B positioned in OC 3 ,

referred to as the fixed reference configuration (assumed to be stress-free). We use the notation
X : D0 ---+ OC 3 for the deformation, which transforms a typical material point X E D0 (with
material coordinates XA, A = 1, 2, 3) to a position x = x(X) E D (with spatial coordinates
Xa, a= 1, 2, 3) in the current configuration, denoted D. Further, let F(X) = ox(X)jfJX be the
deformation gradient and J(X) = detF > 0 the local volume ratio.
Following Flory ( 1961) and Ogden ( 1978), we consider the multiplicative decomposition
(3)
of the deformation gradient into a spherical (dilatational) part J 113 I, and into a unimodular (dis-
tortional) part, denoted by the modified deformation gradient F (detF = 1). We use the sym-
metric right and left Cauchy-Green tensors, denoted C and b respectively, and their modified
counterparts, denoted C and b respectively, associated with F. From eq. (3) we then have
c = FTF = J2f3c, c = FTF, (4)

- - T
b = FFT = j2/3i), b= FF . (5)
Structural and Numerical Models for the (Visco )elastic Response of Arterial Walls... 117
Hyperelastic stress response. In order to describe the hyperelastic stress response of arterial
walls, we employ a set { Aa I a = 1, ... , n} of(second-order)tensors which characterize the wall
structure, and we postulate the existence of a Helmholtz free-energy function tP(C, A1, ... , An)
defined per unit reference volume rather than per unit mass. Note that the free energy depends
on X in addition to C, A 1 , ... , An, but, for simplicity, we do not indicate subsequently this de-
pendency.
Subsequently, we assume the decoupled form
(6)
where the given strictly convex function U, taking on its unique minimum at J = 1, is motivated
mathematically, and the given convex function If/ is responsible for the isochoric elastic response
of the material.
From the Clausius-Planck inequality, standard arguments lead to the well-established expres-
sionS= 28tli(C, A 1 , ... , An)/8C for the second Piola-Kirchhoffstress. Equation (6) then gives
S = Svol +S with (7)
where Svoi and S represent the purely volumetric and isochoric contributions to the second Piola-
Kirchhoff stress S. We shall also require the standard results
8J = ~Jc-1 and (8)

ac 2
from tensor analysis (see, for example, Holzapfel (2000)), where c- 1 = F- 1 F-T is the inverse
of C. The symbol][ has been introduced to denote the fourth-order unit tensor which, in index
notation, has the form (II)IJKL = (81K8JL + 8IL8JK )/2. With these results, eqs. (7)2 and (7)3
become, after some straightforward tensor manipulations and the introduction of the arbitrary
hydrostatic pressure p = dU / dJ as in Holzapfel (2000),
Svol = JpC- 1 , S= J-2/3p: S*. (9)
We used the definitions

1
S* =281f/j8C, r =II- -C
-1
®C (10)
3
for the fictitious second Piola-Kirchhoff stress and the fourth-order projection tensor, respec-
tively. The kinematic quantity JID is introduced with respect to the reference configuration (there-
fore expressed through C), so that (JID : S*) : C = 0 (see Holzapfel (2000)). It is important to
emphasize that in the description of an incompressible material (which an artery is assumed to
be) the function U has no physical meaning any more and the hydrostatic pressure p becomes an
indeterminate Lagrange multiplier.
A Piola transformation of eqs. (9) enables the Cauchy stress tensor u = J- 1 FSFT to be put
in the decoupled form
U = Uvol +U with Uvol =pi, u =I!': u*, (11)

118 G.A. Holzapfel
analogously to eq. (9). Here we used the definition u* = 2J-l F( oWI aC)FT for the fictitious
Cauchy stress and lP' = II - ~I 0 I for the fourth-order projection tensor introduced with respect
to the current configuration, so that (IP': u*) : I= 0.
2.2 Combined Bending, Inflation, Extension and Torsion of a Tube
Basic kinematics. We consider the artery as an incompressible thick-walled cylindrical tube

subjected to various loads. It is known that the load-free configuration, Ores say, in which the
artery is excised from the body and not subjected to any loads is not a stress-free (or strain-free)
reference configuration Do. Thus, the arterial ring springs open when cut in a radial direction. It
appears that Vaishnav and Vossoughi (1983) were the first to publish this finding. The load-free
configuration Ores of an artery arises from certain growth mechanisms of the different layers. For
discussion of growth and remodeling we refer to, for example, Rodriguez et al. ( 1994), Skalak et
al. (1996) and Rachev (1997); for a more recent theoretical work on residual stresses see Ogden
and Schulze-Bauer (2000) and also the lecture notes by Ogden in this course. We assume that
the open sector is the undeformed (stress-free and fixed) reference configuration 0 0 of a circular
cylindrical tube, as depicted in Figure 4. Note that, in general, the residually stressed state is
more complex than considered here. There are also residual stresses in the axial direction of an
arterial segment. For a striking example of this phenomenon the reader is referred to the paper
by Schulze-Bauer et al. (2002).
Thus, in terms of cylindrical polar coordinates (R, G, Z), the geometrical region 0 0 of the
tube is defined by
0 :=:; G :=:; (27r- o:), 0 :S Z :=:; L, (12)
where Ri, R 0 , a and L denote the inner and outer radii, the opening angle and length of the
undeformed (split) tube, respectively.
The deformation x takes 0 0 into the current configuration D. For the considered problem
X = Xp o Xres is the composition of the deformations Xres and Xp, as indicated in Figure 4,
where Xres generates the load-free configuration Ores associated with residual stresses, while
Xp is associated with inflation, axial elongation and torsion of the tube, and leads to the final
configuration n.
In terms of cylindrical polar coordinates (r, (), z), the geometry of the deformed configuration
[! is given by
o:=;z:=;l, (13)
where Ti, r 0 and l denote the inner and outer radii and the length of the deformed tube, respec-
tively.
The deformation x, which is taken to be isochoric, may then be written in the form x =
rer + zez with reference to the (unit) basis vectors {en eo, ez} associated with the cylindrical
polar coordinates (r,B,z), where
p
r= ()=kG+ Z- (14)
L'
Reference
Current
configuration
(stress-free) configmation ~ Mt
X = Xres 0 Xp ~
~ Pi
Pi = 0.0
z
Load-free
N = O.O
configuration
~M, ~ oo
Figure 4. Arterial ring in the (stress-free) reference configuration flo, the load-free configuration flres and
the current configuration fl.
Az is the (constant) axial stretch, the parameter k, defined by k = 27r I (27r - o: ), is a convenient
measure of the tube opening angle in the unstressed configuration, ri is the inner radius in the
deformed configuration and iP is the angle of twist of the tube arising from the torsion. Hence,
the first term kf) in (14)2 represents the deformation from configuration Do to Dres while the
second term ZiP I L describes the influence of the torsion.
In addition to Az, it is convenient to introduce the notations defined by
or R r o(} kr o(} iP
Ar(R) = oR= rkAz' Ae(R) = R o8 = R' 'Y(R) = roz- = r-.
l
(15)
Here Ar(R), Ae(R) and Az are the principal stretches of the deformation associated with the ra-
dial, circumferential and axial directions when there is no twist, while 'Y(R), which is associated
with the twist, represents locally the amount of shear in a (0, z )-plane. Since each of these quan-
120 G.A. Holzapfel
tities depends only on the radius R, the one-dimensional character of the problem is apparent.
When 'Y i- 0, Ar is the principal stretch in the radial direction but .Ao and Az are not then principal
stretches. The condition that the volume is preserved during the deformation is independent of 'Y
and requires simply that
(16)
Note that
(17)
where .Ao i denotes the value of .Ao at the inner surface of the tube. Use of incompressibility
constraint (16) and eqs. (4), (5) enables the deformation gradient F = F and the Cauchy-Green
tensors C = C and b = b to be given in terms of cylindrical polar coordinates (see Holzapfel et
al. (2000a) for more details).
In terms of the parameters k, .Ao i, Az and P, equations (14), (17) define the combined bending,
inflation, axial extension and torsion of a thick-walled tube.
Equilibrium equations. In the absence of body forces the equilibrium equations are
divO' = 0, (18)
where div( •) denotes the spatial divergence of the spatial tensor field ( • ). Note that in cylindrical
polar coordinates (r, (}, z), because of the geometrical and constitutive symmetry, the only non-
trivial component of (18) is
dO'rr + (O'rr- 0'(1(1) = O

(19)
dr r
(see, for example, Ogden (1997)). From this equation and the boundary condition O'rrlr=ro =0
on the outer surface of the tube, the radial Cauchy stress 0' rr may be calculated as
(20)
The internal pressure Pi = -O'rrlr=r; is then obtained in the form
Pi= fro
(0'11(1-
dr
O'rr) --:;:· (21)
r;
This equation will play an important role in the numerical solution of the problem considered.
When the state of deformation is known, expressions for the axial force N and the torsional
couple Mt can be calculated via the definitions
J J
To To
N = 27!' O'zzrdr, Mt = 271' O'ozr 2 dr. (22)

r; r;
In view of the decoupled representation of the Cauchy stress tensor u according to ( 11 h,

we may recast equations (21) and (22)2 by using the decompositions uee = p + Cfee and Urr =
p + (j rr to obtain
Pi= fro
(uee- Urr)
dr
--:;:• Mt = 27r J
ro
Cfezr 2 dr, (23)

r;
where Cfee, Urr denote the isochoric parts ofthe normal components of(Cauchy) stress in the
circumferential and radial directions, while (fez = uez is the shear component of (Cauchy) stress
acting tangentially to the cross-section of the tube.
Use of the additive split ( 11) and eq. (20) enables the axial force N in eq. (22h to be ex-
pressed as
(24)
Reversal of the order of integration in (24) and use of the expression (23 h leads to the general
formula
J
ro
F= 7r (2Cfzz - Cfee - Urr )rdr (25)

r;
for the reduced axial force F = N - r[ 1rPi. This expression for F is very important since it gives
precisely the force that is measured during inflation tests on arteries. A specific form of (25) is
given in eq. (15) of the paper by Chuong and Fung (1983).
The theory described above is designed to capture the deformation behavior in the central
part of a tube so as to exclude end effects. Therefore, axial dependence of the deformation is not
considered. This reflects the typical setting used in experiments (see, for example, Fung et al.
(1979) or Schulze-Bauer et al. (2001 ), amongst others).
Numerical technique to solve the nonlinear boundary-value problem. Here we describe

briefly the numerical technique to solve the problem of bending, axial extension, inflation and
torsion of a thick-walled cylindrical tube.
By assuming a particular state of residual strain (characterized by the parameter k ), the fixed
axial stretch Az and fixed angle of twist P of the tube, the isochoric part of the strain (and hence
the stress) can always be expressed in terms of the two variables A.ei and r, i.e. the circumfer-
ential stretch at the inner surface of the tube and the radius, respectively. Hence, the equation of
equilibrium (23 h may be written in the general form
(26)
r;
where ri is given in terms of A.e i by (17). Since closed-form evaluation of eq. (26) is only pos-
sible for very simple constitutive equations, we employ a Gaussian integration scheme (Hughes
122 G.A. Holzapfel
(2000)), i.e.
(27)
where Wj and rj, (j = 1, ... , n), denote the weights and the Gaussian points, and n is the order
of integration. Equation (27) is, in general, nonlinear in the single unknown >.. 0 i, and, for given
Pi, can be solved for >..o i using, for example, a standard Newton iteration with the initial value
>..oi = 1.0.
Since the deformation is now determined, the torsional couple Mt and the reduced axial
force F follow directly from eqs. (23)z and (25), respectively. This computation is carried out by
employing another Gaussian integration. It turns out that for the considered range of deformations
a three-point integration (n = 3) with the accuracy of order five gives sufficiently accurate
solutions.
2.3 A Multi-layer Model for Arterial Walls
In this section we propose a potential that models each layer of the artery as a fiber-reinforced
composite. The basic idea is to formulate a constitutive model which incorporates some histolog-
ical information. Hence, the material parameters involved may be associated with the histological
structure of arterial walls (i.e. fiber directions), a feature which is not possible with phenomeno-
logical models. The underlying physical background of the described constitutive model leads
to a formulation that is consistent with convexity requirements ensuring mechanically and math-
ematically reliable behavior. The comparative study by Holzapfel et a!. (2000a) reports about
the general lack of convexity detected in several well-established potentials used to describe the
pronounced mechanical behavior of arterial walls. The multi-layer model is based on the theory
of the mechanics of fiber-reinforced composites (Spencer (1984)) and embodies the symmetries
of a cylindrically orthotropic material.
Constitutive model for the arterial layers. Since arteries are composed of (thick-walled)
layers we model each of these layers with a separate strain-energy function. From the engineer-
ing point of view each layer may be considered as a composite reinforced by two families of
(collagen) fibers which are arranged in symmetrical spirals.
We assume that each layer responds with similar mechanical characteristics and we there-
fore use the same form of strain-energy function (but a different set of material parameters) for
each layer. We suggest an additive split of the isochoric strain-energy function Winto a part
Wiso associated with isotropic deformations and a part Waniso associated with anisotropic defor-
mations (Holzapfel and Weizsacker (1998)). Since the (wavy) collagen fibers of arterial walls
are not active at low pressures (they do not store strain energy) we associate Wiso with the me-
chanical response of the non-collagenous matrix material, which we assume to be isotropic. The
resistance to stretch at high pressures is almost entirely due to collagenous fibers (Roach and
Burton (1957)) and this mechanical response is therefore taken to be governed by the anisotropic
function Waniso· Hence, for a representative arterial layer the (two-term) potential is written as
W(C, ao 1, aoz) = Wiso(C) + Waniso(C, ao 1, ao z), (28)

where the families of collagenous fibers are characterized by the two (reference) direction vectors
a 0 i, i = 1, 2, with Ia 0 i I = 1. For a possible identification of collagen fiber orientations recall
Section 1.2.
We now include structure tensors in accordance with the formulation in Section 2.1. Specif-
ically, we incorporate two such tensors, Ai, i = 1, 2, defined as the tensor products a 0 i @ a 0 i.
The integrity basis for the three symmetric second-order tensors C, A 1 , A 2 then consists of the
invariants
(29)
- - -2
14(C,aol) = C: A1, h(C,aol)=C :A1, (30)
16(C, ao2) = C: A2, (31)
(32)
see, for example, Spencer ( 1984), Holzapfel (2000) and the lecture notes by Ogden in this course.
Since the invariants 13 , 19 are constants we may express eq. (28) in the reduced form
(33)
Note that the invariants 14 and 16 are the squares of the stretches in the directions of a 0 1 and
a 0 2 , respectively, so that they are stretch measures for the two families of (collagen) fibers and
therefore have a clear physical interpretation. For simplicity, in order to minimize the number of
material parameters, we consider the reduced form of (33) given by
(34)
The anisotropy then arises only through the invariants 14 and 16 , but this is sufficiently general
to capture the typical features of arterial response.
Finally, the two contributions lj/iso and ljJaniso to the function ljJ must be particularized so as
to fit the material parameters to the experimentally observed response of the arterial layers. We
use the (classical) neo-Hookean model to determine the isotropic response in each layer, and we
write
- - c -
tPiso(h) = 2(h - 3), (35)
where c > 0 is a stress-like material parameter. The strong stiffening effect of each layer observed
at high pressures motivates the use of an exponential function for the description of the strain
energy stored in the collagen fibers, and for this we propose
(36)
where k1 > 0 is a stress-like material parameter and k2 > 0 is a dimensionless parameter. An

appropriate choice of k1 and k2 enables the histologically-based assumption that the collagen
fibers do not influence the mechanical response of the artery in the low pressure domain (Roach
and Burton (1957)) to be modeled.
124 G.A. Holzapfel
All that remains is to determine an expression for the stress, which we provide here in the
Eulerian description. Using (11)3 and the proposed particularizations (35) and (36), we obtain,
after some straightforward manipulations, the explicit isochoric contribution u to the Cauchy
stress tensor, namely
u= cdevb + L 2iPidev(ai 0 ai), (37)

i=4,6
where iP4 = 8iPani so I 814' iP6 = 8iPani so I 816 denote (scalar) response functions and ai = Fao i'
i = 1, 2, the Eulerian counterparts of a 0 i· For a detailed derivation of equation (37) the reader is
referred to the more general constitutive framework described in Gasser and Holzapfel (200 1).
For other applications of the proposed constitutive framework to soft tissue biomechanics, see,
for example, Eberlein et al. (200 1) and Holzapfel (200 1).
Constitutive representation of a healthy young artery. For a healthy young artery, which has
no pathological intimal changes, the innermost layer is not of(solid) mechanical interest, and we
therefore focus attention on modeling the two remaining layers, i.e. the media and the adventitia.
It is then appropriate to model the artery as a two-layer thick-walled tube with residual strains,
as illustrated in Figure 5.
This model uses 6 material parameters, i.e. CM, k1 M, k 2 M for the media and c A, k1 A, k 2 A for
the adventitia. In respect of eqs. (34)-(36) the free-energy functions for the considered two-layer
problem may be written as
- CM - k1 M "" { - 2 }
tPM = 2(h- 3) + 2k ~ exp[k2M(IiM -1) ]-1 ,
2 M i=4,6
CA -
-tPA = -(h- 3) k1 A
+~ "" { - 2 }
~ exp[k2A(IiA -1) ]-1 ,
2 2A._46
t- '
for the media and adventitia, respectively. The constants CM and CA are associated with the non-
collagenous matrix of the material, which describes the isotropic part of the overall response
of the tissue. Note, however, that the matrix material is significantly less stiff than its elastin
fiber constituent. The constants k1 M, k 2 M and k1 A, k2 A are associated with the anisotropic
contribution of collagen to the overall response. The material parameters are constants and do
not depend on the geometry, opening angle or fiber angle.
The invariants, associated with the media M and the adventitia A, are defined by 14 1 = A 1 j :
C and 161 = A 21 : C, j = M, A, and HM is the reference thickness of the media, as illustrated
in Figure 5. The tensors A 11 , A 21, characterizing the structure of the media and adventitia, are
given by
j=M,A, (40)
where, in a cylindrical polar coordinate system, the components of the direction vectors ao 1 j
and ao 2 j have, in matrix notation, the forms
ao21 =[ co~<pj
- Slll<pj
] , j = M,A, (41)
and tpj, j = M, A, are the (mean) angles between the collagen fibers (arranged in symmetrical
spirals) and the circumferential direction in the media and adventitia, as indicated in Figure 5
(compare also with Section 1.2).
Because of the wavy structure of collagen it is regarded as not being able to support com-
pressive stresses. We therefore assume that the fibers contribute to the strain energy in extension
and do not contribute in compression. Hence, in the proposed model the anisotropic terms in the
free-energy functions (38) and (39) should only contribute when the fibers are extended, that is
when 14 j > 1 or 16 i > 1, j = M, A. If one or more of these conditions is not satisfied then
the relevant part of the anisotropic function is omitted from the expressions (38) and (39). If, for
example, 14 A and 16 A are less than or equal to 1, then the response of the adventitia is purely
isotropic.
When these conditions are taken into account, it can be shown that strict local convexity is
guaranteed by the form of the free-energy functions (38) and (39). Strict local convexity means
that the second derivative of tP with respect to the Green-Lagrange strain tensor is positive defi-
nite, with appropriate modifications to account for incompressibility. This fundamental physical
requirement in hyperelasticity ensures that undesirable material instabilities are precluded. It also
induces desirable mathematical features in the governing equations, which are important from the
point of view of numerical computations. It is worthwhile mentioning that some potentials which
have been proposed in the literature are not convex for all possible sets (or any set) of material
parameters. For investigations of convexity for a small number of constitutive models used in
the literature to describe the mechanical response of arteries see the paper by Holzapfel et al.
(2000a).
2.4 Elastic Behavior of a Carotid Artery. Performance of the Model
This section aims to report the performance of the fully three-dimensional constitutive model. In
particular, the anisotropic behavior of a healthy carotid artery under combined bending, inflation,
axial extension and torsion is investigated in more detail.
We use geometrical data from Chuong and Fung (1983) for a carotid artery from a rabbit
(experiment 71 in Fung et al. ( 1979)) and make the assumptions that the media occupies 2/3 of
the arterial wall thickness and that the wall thickness of each layer in the unloaded configuration
(a = 0.0°) is the same as for the case without residual stress (a = 160.0°). The undeformed
length L of the arterial tube was taken to be equal to the value of the inner radius Ri correspond-
ing to a = 0.0°. According to Section 2.2 we make the assumption that the open sector shown
in Figure 5 is stress-free. This simplifying assumption must be regarded as an approximation. In
practice, the opening angles and the stress-free configurations for the separate layers would be
different.
In order to identify the material parameters of the two-layer model for healthy arterial walls,
we fitted the parameters to the experimental data from experiment 71 in Fung et al. (1979) and
used the standard nonlinear Levenberg-Marquardt algorithm. The material parameters obtained
are summarized in Figure 5.
Experimental tests performed by Maltzahn and Warriyar (1984), Yu et al. (1993) and Xie
et al. (1995) indicate that the elastic properties of the media and adventitia are different. Their
results show that the media is much stiffer than the adventitia. In particular, it was found that
in the neighborhood of the reference configuration the mean value of Young's modulus for the
126 G.A. Holzapfel
Materi~ Geometry
r-
= 3.0000 (kPa) H M = 0.26 (mm)
:s
o:l CM
Ql
k1 M = 2.3632 (kPa) I.{'M = 29.0°
~
k2 M = 0.8393 (-)
-- --
o:l
·.c = 0.3000 (kPa) = 0.13 (mm)
Ij
CA HA
k1 A = 0.5620 (kPa) I.{'A = 62.0°

k2A = 0.7112(-)
l R; = 0.7 1 (mm)
R; = 1.43 (mm)
for
for
a= 0.0°
a = 160.0°
Figure 5. Material and geometrical data for a carotid artery from a rabbit in respect of (38) and (39) (see
experiment 71 in Chuong and Fung (1983)).
media, for several pig thoracic aortas, is about an order of magnitude higher than that of the
adventitia (Yu et al. (1993)). For our proposed constitutive model this observation implies that
for these materials the neo-Hookean parameters are such that the ratio cM/cA is typically in the
range of6 to 14. This effectively reduces the number of material parameters, and for definiteness
we therefore set eM = lOcA for purposes of numerical calculation. In general, however, this ratio
depends on the topographical site.
For the functions (38) and (39) the isochoric contribution u to the Cauchy stress tensor u is
obtained from (37). Hence, with the definition (5)2 of the modified left Cauchy-Green tensor b,
the isochoric Cauchy stress components, which are used in eqs. (23)1, (23)2 and (25), are given.
The mechanical response of the carotid artery during bending, inflation, axial extension and
torsion is shown in Figure 6. The internal pressure Pi and the angle of twist cfJ are varied within
the ranges
0 ~ Pi ~ 21.33 (kPa) and - 0.10 ~ cfJ ~ 0.10 (rad). (42)
These loadings are applied at fixed axial stretches of the artery varying between Az 1.5 and =
=
Az 1.9. The solid lines in Figure 6 show numerical results based on a load-free, but not stress-
free, configuration (o: = 160.0°), while the dashed lines are based on a load-free and stress-free
configuration (o: = 0.0°).
(a) (b)
25 ")1 =0.0(·>1
0.18
.....
20
~
~ "- 0.12
;;.
~ 15
~
...
~
~
~
10
~- ~
>.. = 1.7
-=~
A, = 1.6
0.00 >., = L S
5
0 - 0.06
0.3 0.6 0.9 1.2 1.5 0.3 0.6 0.9 1.2 1.5
Inner radius f"1 (mm) Inner rad_ius n (mm)
(c)l4 (d)
p; = 13.33 (kPa)
0.20 A, =l.9 ~
s7
z
~ 0.15
"-
-------
~
e
.. 0.10
: A, = 1.8 - - - - - - - - - - - - - - -
~
=
~ 0.05
>., ~ 1 .7
-----------------
0.00 A, = 1_6- - - - - - - - - - - -
~. - 1.5
- 14
-0.05
-0.15 - 0.10 - 0.05 o.oo 0.05 0.10 0.15 - 0.15 - 0.10 -0.05 0.00 0.05 0.10 0.15
hear -r. (·) Shear ")1 (·)
Figure 6. Deformation behavior of a carotid artery during inflation and torsion using the constitutive model
(38)-(39). Solid lines are numerical results with residual strains included (o: = 160.0°) and
the dashed lines are results without residual strains (o: = 0.0°). Dependence of (a) the internal
pressure Pi and (b) the reduced axial force F on inner radius ri, without shear deformation (/i =
0). Dependence of (c) the torsional couple Mt and (d) the reduced axial force F on the shear ')'i
at fixed internal pressure Pi = 13.33 (kPa). The shaded circles indicate the approximate central
region of the physiological state.
128 G.A. Holzapfel
The internal pressure versus radius behavior is shown in Figure 6(a). The proposed model
(38)-(39) is able to describe the salient features of arterial elasticity, such as the experimentally
observed 'sigma-shaped' form of the pressure/radius relationship; see Figure 2(a) in the paper
by Weizsiicker and Pinto (1988). Note that residual strains have a strong influence on the global
pressure/radius response of the artery. The internal pressure/radius response, of course, does de-
pend on geometry, opening angle and fiber angle, but we have not included here an analysis
of the effect of changes in these quantities. However, our studies have found that in the high-
pressure regime the stress-strain response depends significantly on the fiber angles (as should
be expected). The fiber angles are associated with the stress-free configuration, as indicated in
Figure 5, and we have here assumed that they are the same in the load-free configuration. The
difference in angle between the unstressed and unloaded configurations for the case we consid-
ered goes from (approximately) -3.0° on the inner boundary to +2.7° on the outer boundary
(mean value 0.2°). This approximation has a negligible influence on the analysis.
Figure 6(b) shows that the proposed potential (38)-(39) is also able to model the typical
evolution of the reduced axial force F with inflation (increase of the inner radius) of the artery;
see Figure 2(b) in Weizsiicker and Pinto (1988). This means that F is a decreasing function of
Ti at axial stretches Az less than some value above the physiological stretch and an increasing
function for Az greater than this value.
The response of the artery during torsion at the internal pressure Pi = 13.33 (kPa) is plot-
ted in Figure 6( c)(d). As can be seen from Figure 6( c), the torsional couple Mt increases more
slowly than the shear 'Yi = Pri/l = Pri/ >-zL on the inner boundary increases (i.e. the slope
of the curve decreases). One possible explanation of this interesting phenomenon is as follows:
since the artery is inflated with the internal physiological pressure, the (collagen) fiber reinforce-
ment is activated and the fibers are much stiffer than the matrix material. During torsion from
this state of deformation the nearly inextensible fibers cause the arterial diameter to decrease,
which leads to a reduction in the torsional couple Mt given by eq. (22)2. This realistic diameter-
shrinking behavior of the artery during torsion seems to be a consequence of the considered fiber
reinforcement (orthotropy). However, this effect may also be predicted by a non-convex isotropic
strain-energy function.
In Figure 6(d) the reduced axial force F during torsion is plotted against the shear 'Yi· For
an axial pre-stretch Az = 1.5 the reduction in the inner radius ri due to torsion is about 5.8%
('Yi = 0.119) and 7.8% for Az = 1.9 ('Yi = 0.085). This behavior is in qualitative agreement with
experimental observations presented in Deng et a!. (1994 ).
3 A Structural Model for the Viscoelastic Behavior of Arterial Walls
In this section we present a structural model suitable for predicting reliably the passive time-
dependent three-dimensional stress and deformation states of arterial walls under various loading
conditions. It extends the constitutive framework for the elastic strain response of arterial walls,
as developed in Section 2.3, to the viscoelastic regime (for more details see Holzapfel et a!.
(2002a)).
The reviewed viscoelastic model admits hysteresis loops that are known to be relatively in-
sensitive to strain rate, an essential mechanical feature of arteries of the muscular type. Instead of
modeling the loading and unloading curves by two different laws of elasticity (pseudo-elasticity),
the concept of internal variables is introduced in order to replicate the characteristic dissipative
mechanism of muscular arteries. Evolution equations and associated closed form solutions in
the form of convolution integrals are provided for the internal variables. An efficient update al-
gorithm for the stresses and the associated closed-form expression for the algorithmic elasticity
tensor are briefly discussed. For a theoretical basis of the viscoelastic model at finite strains
and for applications to elastomeric structures see Holzapfel ( 1996) and Holzapfel (2000), Sec-
tion 6.10. A more general constitutive framework for fiber-reinforced composites and associated
computational aspects is provided in Holzapfel and Gasser (2001). In Section 5.2, two represen-
tative numerical examples are designed to demonstrate the good performance and mechanism of
the structural model.
3.1 Constitutive Equations and Internal Dissipation
In order to derive a structural model for the viscoelastic behavior of healthy young arterial walls,
we define a Helmholtz free-energy function iJf, which describes the viscoelastic deformation of
a material point from the reference configuration 0 0 to some current configuration D. The free
energy is assumed to be based on the kinematic decomposition (3) and expressed by the unique
decoupled representation
m
iJf = U(J) +iP(C,A1,A2) + LYa(C,A1,A2,I\,), (43)
a=l
valid over some closed time interval t E [0, T]. The history of the deformation is described by
a set of internal (strain-like) variables, denoted by the second-order tensors r "' 0: = 1, ... 'm.
For the basic idea of internal variables that we use for the description of inelastic processes, see,
for example, Valanis (1972), Lubliner (1990), Simo and Hughes (1998) or Holzapfel (2000),
Section 6.9. Note that all viscoelasticity is assumed to occur purely by isochoric deformations
and all volume changing deformations are forced to be reversible. Hence, the tensorial (history)
r
variables a are akin to C. They are not accessible to direct observation, characterize the current
departure from equilibrium and contribute to the total strain (stress). The viscoelastic behavior is
modeled by o: = 1, ... , m viscoelastic processes with corresponding relaxation (or retardation)
timesTa E (O,oo).
The first two terms on the right hand side of eq. (43) characterize the equilibrium state of
the viscoelastic solid at fixed F as t --+ oo, which is a state of balance, while the third term, the
'dissipative' potentialz:=;=l Y "' characterizes the non-equilibrium state, i.e. the relaxation and
creep behavior. As already mentioned on p. 9, in the description of an incompressible material
(which an artery is assumed to be) the function U is motivated mathematically and serves as a
penalty function within a numerical analysis. For a discussion of the underlying issue the reader is
referred top. 29 in this volume. Representation (43) is well-suited for numerical implementation
(see, for example, Simo et al. (1985), Simo and Taylor (1991) and Holzapfel and Gasser (2001)
amongst others).
Now we particularize the second law of thermodynamics through the Clausius-Planck in-
equality, i.e. Dint = S : C/2 - P ~ 0, where Dint is the internal dissipation (local entropy
130 G.A. Holzapfel
production). By computing the rate of change of tJi and using the chain rule, we find that
( s-Jduc_l_2aw -f2ara)
dJ ac a=l ac
:~c-f
2
a! :r
a=l ar a
>o.
a -
(44)
In deriving (44) we used the properties j = fJJjfJC : C = JC- 1 : C/2 and C= 2(8C/8C) :
C/2.
In order to satisfy Dint ~ 0 for all admissible processes we apply the standard Coleman-Noll
procedure (see Coleman and Noll (1963) and Coleman and Gurtin (1967)). For arbitrary choices
of C, we deduce the constitutive equations for compressible hyperelasticity and a remainder
inequality governing the non-negativeness of the internal dissipation. In particular, the stress
response constitutes an additive split of the second Piola-Kirchhoffstress tensor. We write
f)t]i
+s+L
- m
s = 2 ac = Svol Qa . (45)
a=l
This split is based on the definition
Q = 2 8Ya(C,Al,A2,ra) (46)
a ac
of the isochoric viscoelastic stress contributions Qa, o: = 1, ... , m. The elastic stress contri-
butions Svol and S are according to (9). The stresses Qa may be interpreted as non-equilibrium
stresses in the sense of non-equilibrium thermodynamics, so that the 'dissipative' potential takes
on the form
Ya = j Qa(C*,A1,A2,ra): ~dC, o:=1, ... ,m, (47)

c
(for a detailed exposition of the thermodynamic background see the book by Holzapfel (2000),
which contains further references). The tensor quantities Qa describe purely isochoric stresses,
so that Qa : C = 0, o: = 1, ... , m. The internal dissipation is given by
~OYa -'-
Dint = - L....J ---=- : r a ~ 0. (48)
a=l ar a
r
Note that for arbitrary elastic processes a = 0, and hence the internal dissipation Dint is zero
(the material is considered to be elastic).
This constitutive framework, which is based on the concept of internal variables, is now used
for the representation of viscoelastic stress responses of arterial layers.
3.2 Decoupled Stress Response

In this section we are concerned with the viscoelastic stress response of healthy young arteries of
the muscular type at finite strains.
Viscoelastic stress response in the material description. As for healthy young arteries of
the elastic type (recall pp. 16, 17) we model the artery as a two-layer thick-walled tube, i.e.
the media and the adventitia, which make the significant contributions to the solid mechanical
properties of the wall. Hence, for each arterial layer the stress response is derived from the as-
sociated Helmholtz free-energy function. According to the constitutive framework discussed in
Section 3.1 we arrive at the (second Piola-Kirchhoft) stress response
m
SM = (JpC- 1)M + L(Sa + L Qaa)M,
a=1,4,6 a=l
(49)
m
SA= (JpC- 1 )A + L:Csa + L Qaa)A
a=1,4,6 a=l
for the media and adventitia, respectively (compare with the mathematical structure (45)2). How-
ever, S, Qa are replaced by Sa, Qaa signifying the a-th constituent of the arterial layer (a= 1
is associated with the non-collagenous matrix material and a = 4, 6 with the two families of
collagen fibers). We recall that the index a stands for the number of viscoelastic processes. Note
that for the case a = 1 isotropy is recovered as a special case.
Considering the particularization (38) and (39) of the free energy and the decoupled repre-
sentation of the kinematics according to Section 2.1, we may express the isochoric elastic stress
contributions (eq. (9) 2 with definition (lOh in the convenient form
(a= 1, 4, 6; no summation), (50)
which is used in relations (49). We introduced the useful definition
a=1,4,6 (51)
of the stress functions '1/Ja, which are affected by the special choice of ill (the material). In addi-
tion, in eq. (50) we have introduced the definitions
D = ala a=1,4,6 (52)

a f)C '
of kinematic quantities, i.e. the second-order tensors Da, a = 1, 4, 6. From eq. (52) we may
conclude, using definitions (29h, (30h, (31h, that D 1 =I, D4 = A 1 and D6 = A 2 . With the
given functions (38) and (39) it is also straightforward to particularize eq. (51) in order to obtain
'l/J1 = c/2, (53)
'l/J4 = k1(I4 -1)exp [k2(l4 -1) 2], (54)
'l/J6 = k1(I6 -1)exp [k2(l6 -1) 2] (55)
As seen from definition (50) (with eqs. (53)-( 55)) the stress response consists of purely isotropic
contributions sl due to the matrix material, and anisotropic contributions 84, 86 due to the two
families of fibers, which characterize decoupled stresses (associated only with the fibers).
The non-equilibrium states of the arterial layers are associated with the additional tensor
variables Qa a• for a = 1, ... , m and a = 1, 4, 6. They are zero at a state of thermodynamic
132 G.A. Holzapfel
equilibrium, which implies that the anisotropic material responds perfectly elastically. The non-
equilibrium stresses are governed by complementary equations of evolution, as discussed in the
following section.
Evolution equations for the non-equilibrium stresses. In order to determine how a vis-
coelastic process in an artery evolves, we have to postulate additional equations governing the
non-equilibrium stresses.
Hysteresis loops of arterial tissues are known to be not very sensitive to strain rates over
several decades (see, for example, the study by Tanaka and Fung (1974) performed for various
arteries of dogs). This is also true for other types of biological soft tissues such as articular car-
tilage (Woo eta!. (1979)) or the mesentery (Chen and Fung (1973)). Hence, we have to select a
rheological model which considers this characteristic feature. Classical mechanical devices such
as the Maxwell model (spring in series with a dashpot ), the Kelvin- Voigt model (spring in parallel
with a dashpot) or a device of the 'standard solid' type, which is a free spring on one end and one
Maxwell element arranged in parallel, are not able to represent the typical viscoelastic behavior
of soft tissues. The damping mechanisms of these devices are strongly frequency dependent and
are not suitable candidates for formulating meaningful evolution equations. However, a mechani-
cal device which is composed of a number of springs and dashpots gives the required viscoelastic
behavior (see, for example, Fung (1993), Section 7.6, for more details and references).
For this reason we extend the attractive one-dimensional generalized Maxwell model to the
three-dimensional region. The generalized Maxwell model may be seen as a mechanical device
with a free spring on one end and an arbitrary number m of Maxwell elements arranged in par-
allel (see Holzapfel (2000), Section 6.10; in particular, see Example 6.10). The more Maxwell
elements and associated (different) relaxation times used the nearer is the response to constant
damping over a wide frequency spectrum.
Hence, for each of the (isochoric) non-equilibrium stresses, separately for each a and con-
stituent a of the arterial layer, we formulate an evolution equation. We assume the set of linear
differential equations
Qaa (3oos
Q. aa+--= (a= 1, 4, 6; no summation),
aaa•
Taa
anda=l, ... ,m) (56)
valid for some semi-open time interval t E (0, T]. The initial conditions (56)2 ensure that the
reference configuration has no viscoelastic stress contribution. The constants f3':a E [0, oo) in-
troduced are given so-called free-energy factors, which are non-dimensional and associated with
the relaxation times r a a E (0, oo), which describe the rate of decay of the stress and strain in a
viscoelastic process.
Closed-form solutions of the linear equations (56)1 may be represented by the simple convo-
lution integrals
J
t=T
Qaa = exp[-(T- t)/raa]f3~aSa(t)dt , (57)

t=O+
for a = 1, 4, 6 and o: = 1, ... , m. The typical features of anisotropic arterial response in the large
strain domain are now described completely by the constitutive equations (49), with expressions
(50)-( 55) and (57).
Algorithmic stress tensor at time tnH. In the following we review an efficient update al-
gorithm for the stresses suitable for the numerical implementation in a finite element program.
The update algorithm, which is basically concerned with the numerical integration of the convo-
lution integrals (57), goes back to Herrmann and Peterson (1968) and Taylor et al. (1970). For a
detailed exploitation of the time integration algorithm (for a = 1) and for additional references
see Holzapfel (2000), pp. 290-293.
We consider some closed time interval t E [o+, T] of interest and focus attention on a typical
closed time sub-interval [tn, tn+ 1 ], with Llt = tn+ 1 - tn characterizing the associated time incre-
ment. Assume now that at certain times tn and tnH all relevant kinematic quantities are given.
In addition to that, the stress Sn at tn is also specified uniquely via the associated constitutive
equation serving as an 'initial' data base. All that remains is to compute the so-called algorithmic
stress tensor Sn+ 1 at tn+l according to
m
Sn+l = (Svol +S+ L Qa)n+1· (58)
a=1
The first two elastic stress contributions Svol nH and Sn+ 1 can be computed simply from the
given strain measures at tn+l· The third term QanH in (58), which is based on convolution
integrals (57), is responsible for the viscoelastic stress contribution and remains to be evaluated.
Using the mid-point rule we arrive at a second-order accurate recurrence update formula for the
non-equilibrium stresses, namely
Qan+1 = 1-lan + L(Qaa)n+1 ' o: = 1, ... ,m, (59)

a=1,4,6
with the definition
1-lan = L exp(-Llt/2Taa)[exp(-Llt/2Taa)(Qaa)n- JJ~a(Sa)n] (60)

a=1,4,6
of the (algorithmic) history term 1-la n, o: = 1, ... , m, at tn. In addition to the history term we
have used the definitions
(61)
of the non-equilibrium stresses Qa a at tn+ 1 and the so-called viscoelastic factors c5a a, for each
o: and a. Note that the update algorithm is formulated in terms of material coordinates, and hence
the objectivity requirement based on an Euclidean transformation is trivially satisfied.
3.3 Elasticity Tensor in the Material Description
The use of the viscoelastic model described above within a finite element context, with which
it is desired to combined a Newton-type solution method, requires knowledge of the linearized
134 G.A. Holzapfel
constitutive equation (for specific solution methods see Section 4.4 in this volume). In general,
<C = 28S I ac is defined to be the fourth-order elasticity tensor (tangent moduli) in the material
description. It is the gradient of the nonlinear tensor-valued (tensor) functionS, which possesses
minor and major symmetries (Holzapfel (2000)).
We now restrict our attention to the purely isochoric (elastic and viscoelastic) contribution to
<C, which we wish to denote by Cv. Given the structure of the decoupled stress relation (45), the
associated isochoric part of the elasticity tensor at time tn+ 1 may be written in the form
c~+1 = [2: (1 + 8a)CalnH, (62)

a=1,4,6
where c~+l is referred to as the algorithmic elasticity tensor in the material description at t n+ 1·
In eq. (62) each factor 8a is determined as the sum of the associated m viscoelastic factors 8a a,
a= 1, ... , m, given through eq. (61)2. Note that the viscoelastic contribution to the algorithmic
elasticity tensor c~+l at tn+1 may be expressed in terms of the viscoelastic factors 8a a. govern-
ing the time-dependent part, and (Ca) n+ 1 at tn+ 1 , which is associated with the isochoric elastic
response as t -+ oo. For a detailed derivation of (62), in particular of Ca, the reader is referred
to the Appendix of the paper by Holzapfel and Gasser (2001). Here, only the explicit result for
the isochoric part Ca is presented, i.e.
with the abbreviations
Tr(•) = (•): c (64)
of the fourth-order modified projection tensor JlD and the trace Tr( • ), respectively. In eq. (64h
we used the definition -C- 18 c- 1 of the fourth-order tensor ac- 1j8C, for convenience. The
symbol 8 has been introduced to denote the tensor product according to the rule
-1 c-1) 1 ac- 1
- (c 8 ABCD = - 2(c-1 c-1 c-1 c-1)
AC BD + AD BC = oCcD.
AB (65)
Recall that the kinematic quantity JlD is given by eq. (1 0) 2 , and the stress contributions Sa and
s: by eq. (50). In addition, in eq. (63)2 we have used the definition c: = 2J- 4 13as:;ac,
a = 1, 4, 6 of the fourth-order fictitious elasticity tensors in the material description. Their closed
form expressions are
c: = 4J- ! 2: 'lf'abDa IZi Db,

4 3 (a= 1, 4, 6; no summation), (66)
a=1,4,6
in which we have introduced the useful definitions
a, b = 1, 4, 6 (67)
of the elasticity functions 'if; ab, which depend on the special choice oftJr. The second-order tensors
Da are introduced in eq. (52). With the given functions (38) and (39) it is straightforward to
particularize eq. (67). The non-vanishing scalars are
'lj;44 = k1 [I+ 2k2(l4- 1) 2] exp [k2(l4- 1) 2] , (68)
'l/J66 = k1 [1 + 2k2(l6- 1) 2] exp [k2(l6- 1) 2] (69)
Note that for the case of thermodynamic equilibrium in which t ---+ oo, the non-equilibrium
stresses vanish and the isochoric part c~+l of the elasticity tensor given by eq. (62)1 reduces to
C. General finite anisotropic elasticity is recovered with
C= Lea, (70)
a=1,4,6
and the constitutive framework for the description of the elastic behavior of arterial walls holds,
as presented in Section 2.
4 Mixed Finite Element Formulation for Incompressible Finite Elasticity

Various types of biological soft tissues show a very high resistance to volumetric changes com-
pared with that to isochoric changes. Hence, due to the nearly incompressible or incompressible
deformation behavior of soft tissues a very careful numerical treatment is required. A conven-
tional displacement-based method cannot be applied directly to these types of kinematic con-
straint problems, because 'locking' and 'checkerboard' phenomena will occur. A mixed finite
element formulation is to be used.
In this section we describe one particular simple and well-established mixed finite element
method in detail, which turns out to be very efficient within the framework of incompressible
finite elasticity. We prefer to use the Eulerian description and define the tensor and matrix prob-
lems for fully nonlinear elastodynamics. Without loss of generality, we choose a Lagrangian
function which does not depend on structural tensors as introduced in Section 2.1. We start with
Hamilton's principle, as the counterpart of the principle of stationary potential energy in dynam-
ics, and continue with a brief review of the three-field variational principle by Simo et al. ( 1985),
which takes account of nearly incompressible response. Spatial discretization of the continuous
problem gives three Euler-Lagrange equations in weak form. Thereby, the volumetric variables
are eliminated at the element level leading to the quite robust QI/ PO-element, which circum-
vents numerical difficulties that arise from the overstiffening of the system associated with the
analysis of isochoric constitutive responses of arterial walls. By a consistent linearization pro-
cess we arrive at the stiffness matrices suitable for implementation in a finite element program.
The section concludes by providing some insights into solution methods for nonlinear dynamic
problems.
4.1 Variational Principles

Hamilton's variational principle. The important Hamilton 's variational principle represents
the laws of dynamics of (discrete and) continuous systems in concise form. It is equivalent to
136 G.A. Holzapfel
the principle of virtual work and the Euler-Lagrange equations. As a particularization for static
problems, the principle of stationary potential energy may be obtained. Hamilton's principle is
presented by the stationary condition
t,
6 j L(u, ti)dt = 0, L(u, ti) = II(u)- K(ti), (71)

to
where u(x, t) = u(X, t) denotes the displacement vector field of a typical particle of a continuum
body l3 positioned in IR 3 . It relates its position X in the undeformed configuration to its position x
in the deformed configuration at timet. The displacement vector field is regarded as continuously
differentiable with respect to position and time. The velocity field, denoted by ti(x, t) = ti(X, t),
is the material time derivative of u. The scalar-valued functionals II and K denote the total
potential energy and the kinetic energy of the moving body, respectively. The difference between
II and K, called the Lagrangian function L (in the literature sometimes introduced as - L ), is
integrated with respect to timet over a closed time interval t E [t 0 , tl] (to and t 1 are two arbitrary
instants of time). We now assume the restriction that at the times t 0 and t 1 , the variation of the
displacement vector field, denoted by 6u, vanishes at all points of the continuum body, i.e.
(72)
Note that 6u is regarded as a function of position and time.

For further considerations we assume that the reference configuration Do of the continuum
body l3 is bounded by a reference boundary surface 8D0 . We agree that this boundary surface
is partitioned into two parts, 8Dou and 8Dou, on which Dirichlet and von Neumann boundary
conditions will be specified. The two parts are non-overlapping so that
oflo = oflou U 8Dou with oflou n 8Dou = 0. (73)
Subsequently, we characterize the total potential energy II as the sum of the internal and external
potential energies, IIint and IIext, i.e.
(74)
IIint(u) = J .P(C(u))dV, IIext(u)=- J B·udV- J T·udS, (75)

!Jo !Jo 8!1o"
where lJi represents the strain-energy function defined per unit reference volume. It is expressed
as a function of the six components CAB of the right Cauchy-Green tensor C. The infinitesimal
volume element dV and the infinitesimal surface element dS are defined in the reference con-
figuration. In addition, we assume here that the loads do not depend on the motion of the body
(IIext is a linear function ofu). The reference body force B = B(X, t) is referred to the referential
position X (force measured per unit reference volume), while T denotes the prescribed (given)
first Piola-Kirchhofftraction vector (force measured per unit reference surface area) specified on
the part 8D0 u c 8D0 of the boundary surface.
The kinetic energy of a moving continuum body is defined to be
K(ti) =I ~pau2 dV,

S?o
(76)
where p 0 (X) E Q0 is the (time-independent) reference mass density.

As we known from the stationary energy principle, the vanishing variation of L with the
imposed restrictions (72) gives the principle ofvirtual work for all tl"u which are zero on the part
8flou c ana of the boundary surface where the displacement field is prescribed throughout the
entire closed time interval t E [to, t1].
Simo-Taylor-Pister variational principle. The principle of virtual work is not the appropriate
variational approach for invoking kinematic constraint conditions such as incompressibility ( J =
1) frequently arise in soft tissue biomechanics. Within variational methods, constraint conditions
may be considered by introducing Lagrange multipliers in the associated functional (see Washizu
(1982)).
Here we recall the well-established variational principle proposed by Simo et al. (1985),
which will result in a simple but efficient finite element formulation. It takes account of nearly
incompressible response and is based on the kinematic assumption
(77)
Here, J is a kinematic field variable which enforces the kinematic constraint J = J(u) (J > 0 is
the local volume ratio), and F = J- 113 F is the modified deformation gradient. The deformation
measure is given by C = FT F = J 213 C(u), where C = J- 2 13 C is the modified right Cauchy-
Green tensor as introduced in Section 2.1. Thus, for the strain-energy function IJi we may write
lfi = U(J) + iP(C(u)). (78)
The three-field variational principle of Simo et al. (1985), which is of Hu- Washizu 's type,
treats the displacement and pressure fields u and p, and the kinematic field variable J as inde-
pendent variables. Hence, we extend the Lagrangian function (71 )z to the Lagrangian function
LsTP, which takes on the form
LsTP(u,p, ]) =I [U(J) + p(J(u) - ]) + iP(C(u))]dV + Jiext(u)- K(ti), (79)

S?o
where eqs. (74), (75)1 and (78) have been used. The constraint condition J = J is enforced by
p. A suitable candidate for the function U is
(80)
with U(l) = 0. Note that for the incompressible case the (positive) parameter J£ > 0 serves as
a user-specified penalty parameter which has no physical relevance. Since J£ > 0, dU 2 ( ]) / d 2 J
is positive, and hence U is convex. It is worthwhile mentioning that the Lagrangian function
(79) is often augmented by a term of the form Jiaug(J; .>..) = fno .Ah(J)dV, where h(J) is a
138 G.A. Holzapfel
continuously differentiable function that satisfies h(l) = 0 (for example, h = (J- 1)), and
A is a Lagrange multiplier. According to Simo and Taylor (1991), A may be determined by a
standard update procedure of the form A {= A + "'h( J) until the magnitude of h is less than a
given tolerance of accuracy. The augmented term enforces the incompressibility constraint and
prevents the global stiffness matrix from becoming increasingly ill-conditioned for increasing"''
a problem known from the penalty method.
In addition to the virtual displacement field b"u we now introduce arbitrary smooth functions
b"p and b"J, which may be interpreted as the virtual pressure field and the virtual volume change,
respectively. Hence, the necessary stationary conditions for the function LsTP with respect to
the three field variables (u, p, ]) are evaluated separately. We require the stationary conditions
I I I
tl tl tl
Dou LsTpdt = 0, Dop LsTPdt = 0, D 0J LsTPdt = 0, (81)

to to to
for all clu satisfying tlu = o on the part 8Do u of the boundary surface 8Do throughout the entire
time interval t E [to, h], for all clu that vanish at times t 0 and t 1 at all points of the body (compare
with eq. (72)), and all tlp, 6].
In eqs. (81) we have introduced the concept of directional derivative and used the equivalent
representation of the first variation and the directional derivative of a functional, .P say, according
to
(82)
where D(•) is the Gateaux operator and sis a scalar parameter. We say that DJuP(x) is the
directional derivative of <P at given x (fixed) in the direction of the virtual displacement field tlu.
By means of the chain rule we immediately find from definition (82) that D 00 P = (8<Pj8x) · tlu.
Next, we differentiate LsTP, as provided in eq. (79), with respect to changes in u. By using
the commutative property that the order of the directional derivative and the definite integral is
interchangeable, then eq. (8h and the analogue ofeq. (7)3, we find from (8lh that
II
tl
to no
(JpC- 1 + S) : ~DauCdV dt +I tl
to
Daullext(u)dt- Ili
to
DauK(u)dt = 0, (83)
where DauC characterizes the directional derivative ofC in the direction of b"u, and S represents
the purely isochoric contribution to the second Piola-Kirchhoff stress S. From expression (75)2
we find that the second term in eq. (83) has the form
I =-I I II
tl li tl
D 6 unext(u)dt B · b"udVdt- T · b"udSdt. (84)

to to no to 8no <T
Next, we specify the third term in eq. (83) by using the kinetic energy (76) and interchanging
the order of the directional derivative and the definite integral. Subsequently, integration by parts
gives
tl tl h
-j DauK(ll)dt = - j Dau j ~Poll 2 dVdt = - jj Poll· DaulldVdt

to to Slo to Slo
J JJ
tl
= - Poll· DauudV + Poll· DauudV dt. (85)

Slo to to Slo
With the equivalent notation D au u = llu and restriction (72) we deduce from (85)3 that
tl h
- j DauK(ll)dt j j = Poll· lludV dt. (86)

to to Slo
Finally, with eqs. (84), (86) and the equivalent notation DauC = 6C we may rewrite eq. (83) in
the variational form
j [(JpC- 1 + S) : ~6C- (B- Poll)· llu]dV- j T · lludS = 0, (87)

~ 8~~
with the condition llu = o imposed on the boundary surface aflou C an0 . Note that eq. (87)
holds for each t E [to, h].
A straightforward differentiation of LsTP with respect to changes in the field variables p and
J gives the weak enforcement of the equivalence between J and J, and the expression for the
volumetric changes. Hence, from eqs. (81 )2 and (81 )3 we find the remaining two Euler-Lagrange
equations in weak form
Slo
J (J(u)- ])JpdV = 0, (88)
Slo
J( dU(!) - ) 6]dV = 0
dJ p '
(89)
which hold for each t E [to, t!]. Note that for arbitrary 6p, the variational equation (88) results in
the local form J = J of the kinematic constraint condition. For arbitrary 6], eq. (89) gives the
standard constitutive equation p = dU/ dJ.
For subsequent use it is beneficial to express the stationary condition (87) in terms of spatial
coordinates. Before proceeding we assume that region flo moves to a new region fl occupied
by T3 at a subsequent time t > 0. It is bounded by a current boundary surface an, which is
decomposed into disjoint parts so that an = aflu U aflu with anu n anu = 0. According to,
for example, Holzapfel (2000), Sections 8.2, 4.3, we transform now the material fields 6C and B
into the associated spatial fields lle and b. This yields,
1 T
2Jc = F JeF, B= Jb, (90)
140 G .A. Holzapfel
where b denotes the body force with respect to the current position x (force measured per unit
current volume), and
6e = 21 (gradT 6u + grad6u) = sym(grad6u), (91)
is the first variation of the Euler-Almansi strain tensor e. Note that the symbol b should not be
confused with the strain tensor introduced in eq. (5h. In relation (91 ), grad( •) denotes the spatial
gradient of ( •) (lowercase g), which is the derivative of ( •) with respect to the current position
x, at a fixed timet, and the notation sym( •) is used to indicate the symmetric part of (• ).
The change in volume between the current and the reference configuration and the relation
between the reference mass density and the spatial mass density, denoted by p(x, t), at timet, is
given by
dv = J(X, t)dV, po(X) = J(X, t)p(x, t). (92)
Here, dv defines the infinitesimal volume element in the current configuration.

With relations (90)-(92) we deduce from (87) the corresponding function in terms of spatial
quantities, i.e.
/[(pi+ iT): 6e- (b- pu) · 6u]dv- t · 6uds = 0, j (93)

a a~
where a'= J- 1 FSFT defines the purely isochoric stress contribution to the Cauchy stress tensor
u. The virtual displacement field 6u is here defined on the current configuration (we require that
6u vanishes on 8Du)· The prescribed Cauchy traction vector (force measured per unit current
surface areas), denoted by t, is specified on the portion anu an.
c
For subsequent use it is useful to rephrase eqs. (87) and (93) in the more convenient form
(94)
in which we have introduced the definitions (in the Lagrangian and Eulerian descriptions)
6Wint = j (pl +iT) : 6edv = j (JpC- 1 + S) : ~6CdV, (95)

a
J J
Do
6Wext = 6Wtt + 6W~xt, 6Wbxt = b · 6udv = B · 6udV, (96)

a Do
6W~xt = J t · 6uds = JT· 6udS, (97)

BD" BDou
6Wkin = j pu · 6udv = j p0 u · 6udV, (98)

a Do
of the modified internal virtual work 6Wint, the external virtual work 6Wext = -D 0uiiext(u)
and the kinetic virtual work 6Wkin. We call the nonlinear variational equation (93), or (94), the
modified virtual work equation.
It is important to note that the second-order differential equations themselves require addi-
tional data in the form of initial conditions. We agree that the initial configuration, i.e. a region
at initial time t = 0, coincides with the reference configuration; hence, the reference time is at
t = 0. The displacement field ult=O and the velocity field 1ilt=O at the initial time t = 0 are
specified in the weak forms
I u(x, t)lt=o · 8udv =I u0 (X) · 8udv, )

Q
Q
I u(x, t)lt=O 8udv
0
=I
Q
Q
Uo(X) 8udv,
0
(99)
where ( • ) 0 denotes a prescribed function in Do. This set of initial conditions together with
eqs. (88), (89) and (93) characterizes the weak form (or variational form) of the nonlinear initial
boundary-value problem.
If the local forms J = J and p = dU / d] are satisfied (compare with eqs. (88) and (89) ),
then p is identified as the hydrostatic pressure and eq. (93) characterizes precisely the virtual
work equation in the spatial description. Thereby, the term pi determines the purely volumetric
contribution u vol to the Cauchy stress tensor u. Hence, for vanishing acceleration ii, the internal
virtual work, then governed by c>wint = fa u : 8edv, with u = (dU/dJ)I + u, equals the
external virtual work c>wext, so that 8W = 0.
4.2 Finite Element Discretization
The aim of this section is to spatially discretize (approximate) the described continuous problem
of solid mechanics (the time remains continuous at this point). In particular, we derive the discrete
versions of the three nonlinear Euler-Lagrange equations (88), (89) and (93) in weak form. For
establishing the discrete quantities we prefer subsequently to use the Eulerian description.
We subdivide the current domain{} and the portion 8{}u c 8{} of the boundary surface{){}
of a continuum body B into ne finite elements with domain ne and nb element surfaces with
domain 8Du b on which traction vectors are specified. We write
nb
anu ~A anub, (100)

b=1
where A denotes an operator which assembles the elements. Subsequently, the subscript (• )e
is an index (running between 1 and ne) which will refer to a typical finite element De, and the
subscript (. )b (running between 1 and nb) which will refer to a typical element surface anu b·
Discrete kinematics. We prefer to use isoparametric elements and interpolate the geometry
of the reference and current configurations according to
2: N1(f,)X1,
nnode
2: N1(f,)x1(t),
nnode
Xh = Xh = (101)
1=1 1=1
142 G .A. Holzapfel
Parent domain Current configuration
(1,1,1)
(-1,-1,-1)
Figure 7. Three-dimensional element in the current configuration and the parent domain.
where the subscript ( •) h indicates the discrete (finite-dimensional) counterpart to quantity ( •).
The reference and current positions of the element node I are denoted by XI and XI, respectively.
The subscript ( •) I is an index running between 1 and the total number of element nodes, denoted
e
by nnocte· In eqs. (101), = {6,~2,6} E Do are the local element coordinates (the natural
coordinates), where Do = {e E IR3 1( -1, 1) x ( -1, 1) x (-1, 1)} characterize the domain of
the parent element, i.e. a biunit cube, as illustrated in Figure 7. The isoparametric interpolation
(or shape) function associated with node I, is denoted by NJ(e) and defined in Do (for a more
detailed explanation of these standard concepts, see, for example, Hughes (2000)). Now we shall
assume that the interpolation functions are tri-linear and expressed in the form N I (e) = (1 + i
~u6)(1 + ~u6)(1 + 616), I = 1, ... , nnode = 8.
With eqs. (101) it is straightforward to compute the discrete displacement uh = xh- Xh and
the acceleration iih, i.e.
= L: 2: NI(e)ui(t),
nnode nno de
uh NI(e)ui(t), uh = (102)
I=1 I=1
where the nodal displacements and the nodal accelerations are denoted by ui(t) = xi(t) -XI
and iii(t) = x1 (t), respectively. As can be seen from eqs. (101) and (102), a specific element is
characterized by the positions and the degrees of freedom associated with the element nodes I
attached to it.
In addition, the interpolations of the virtual displacement 8u, and the related spatial gradient
grad8u and spatial divergence div8u give the discrete forms
8uh = L NI(t_)8ui, (103)

I=l
L
nnode
grad8uh = 8ui Q9 V xNI(t_), (104)

I=l
L 8ui · V xNI(t_),
nuode
div8uh = (105)
I=l
where the standard expressions
(106)
are to be used. Therein, V (.) N I denotes the gradient of the scalar function N I with respect to
the coordinates ( •) and jh is the Jacobian operator, which transforms the gradient V t;NI to
the gradient V xNI. The utilization of the same interpolation functions to discretize u and 8u is
called the Bubnov-Galerkin (finite element) method.
According to (91 ), the discretization of the variation of the Euler-Almansi strain tensor 8e =
sym(grad8u) within a typical finite element domain may be found with representation (1 04) as
l
L (8ui Q9 V xNI + V xNI Q9 8ui).
nnode
8eh = sym(grad8uh) = "2 (1 07)

I=l
In addition, for the independent volumetric variables J, p, and their variations 8], 8p, we use
the same constant (discontinuous) interpolation functions over a given element domain without
having to satisfY continuity across the element boundaries (Simo et al. (1985)). Thus, we write
(108)
where the symbol (i) denotes the constant interpolation function. This type of mixed formulation
is known as the mean dilatation technique, leading to the QI/ PO-element, a procedure which
goes back to Nagtegaal et al. (1974). This approach may be regarded as the nonlinear extension
of the B-bar method, as proposed by Hughes (2000). It turns out that this approach leads to
a finite element formulation which prevents 'locking' phenomena associated with the volume-
preserving condition on the motion. For practical use the QI/ PO-element is quite robust and is
therefore implemented in several commercially available finite element programs.
Discrete variational equations. We refer to eq. (94) and study the three terms 8Wint, 8wext
and 8Wkin, i.e. (95)-(98), in more detail. For the three contributions to the modified virtual work
8W we have the discrete forms
ne
8Wint ~ L 8W~nt' (109)
e=l
144 G.A. Holzapfel
ne nb
8wext ~ L 8Wt~t + L 8W:~t' (110)
e=1 b=1
ne
8Wkin ~ L 8Wekin' (111)
e=1
which constitute the basis for the finite element formulation.
Just for clarity, we consider subsequently the virtual work contributions caused by a typical
finite element with domain fle and surface 8fla b (on which traction vectors are specified). We
start with the modified internal virtual work 8W~nt, which may be derived from definition (95) 1 .
By means of the interpolations (107)2 and (108) 1 , the basic properties of double contraction and
the symmetry of the Cauchy stress tensor, we obtain
8W~nt = j (phi+ ch): 8ehdv = nfe j (pi+ ch): (8u1 0 VxNI)dv

~ ~1~
nfe 8u1 ·
/=1
J(pi+ Cfh)V xNJdV. (112)
The Cauchy stress tensor Cfh is a function of llh through the respective constitutive relation.
By means of the discrete equation (1 03), the external virtual work for a finite element fle
with surface 8fla b results from definitions (96) and (97) as
(113)
(114)
with the given boundary conditions b and t. Note that for the term 8W~~t the interpolation
functions are to be used for an element, which is reduced by one space dimension. Thereby,
the interpolation functions are denoted by N 1, I = 1, ... , bnode.
We now generate the kinetic virtual work 8Wekin within a typical finite element domain by
substitution of the interpolations (102) 2 and (103) into definition (98)1, causing
(115)
Finally, the initial conditions, which are the given displacement and velocity fields, are ex-
pressed in the discrete forms as
llJit=O = Uo(XJ), I = 1, ... , nnode, }

(116)
uiit=O = tio(XJ), I = 1, ... , nnode
(rather than enforcing (99)), where XI characterizes the referential position of element node I.
The two equations (88) and (89) have to be satisfied locally, since the interpolation functions
used for the (independent) volumetric variables J and p are constant (discontinuous). Therefore,
with the four quantities provided in eq. (1 08), we find that the discretizations for the volumetric
variables have the forms
;;
J =I
1
dV
J JhdV = Ve'
Ve (117)
floe floe
j5 =
1 J q) v
I floe dV floe
dU
dJ
d = dU q) I
dJ }-=ve/Ve
(118)
where Ve and Ve are element volumes in the current and reference configurations, respectively.
In In fn
They are defined by J&Qe JhdV = J&e dv = Ve and J&Qe dV = Ve. Hence, both volumetric
variables can be eliminated at the _element level (static condensation), and the expression (118) 2
for the constant pressure j5 (with J determined in (117)2) is therefore used in eq. (112)3.
Finite element discretization in matrix notation; Semidiscrete nonlinear equations of mo-

tion. The discrete equations (112)3, (113)2, (114)2, (115)2, which are given in tensorial forms,
are now recast in the more standard matrix notation.
The analogues of eqs. (102)2 and (103) in matrix form are uh = L,?;.ot NI(f.)ui(t) and
8uh = L,~;_ot NI(f.)8ui(t), where UI and 8ui denote 3 x 1 column matrices of the form
(119)
They contain the components {ii 1 , u2 , u3 } I of the acceleration vector iii and the components
{8u 1 , 8u2, 8u 3 } I of the virtual displacement vector 8ui for node I.
Subsequently, we represent the first variation of the symmetric Euler-Almansi strain tensor
8eh within a typical finite element domain by the 6 x 1 column matrix, denoted by 8eh. It is of
the form
(120)
where, for the three-dimensional case, the (six) independent components of the tensor 8eh form
the entries of the column matrix 8eh. With this convention we may express the discrete equation
146 G.A. Holzapfel
(1 07)2 in the standard matrix form
aNI
0 0
axl
aNI
0 0
ax2
aNI
0 0
nnode
ax3
8eh = 2:=
I=l
BI8u~, BI= aNI aNI
0
(121)
ax2 axl
aNI aNI
0
ax3 ax2
aNI aNI
0
ax3 axl
The 6 x 3 matrix B I is associated with node I, consists of the derivatives of the interpolation
functions NI with respect to the spatial (current) coordinates X a, and is, therefore, configuration
dependent.
By analogy with eq. (120) the isochoric contribution uh to the symmetric Cauchy stress
tensor of a typical element is given by the 6 x 1 column matrix ([ h of the form
(122)
where the (six) independent components of the tensor uh form the entries of the column matrix
'fh· The second-order unit tensor I is represented by the matrix I.
By means of(112h and the interpolations (108)1 and (121)1 we find from the approximation
(109) the modified internal virtual work owint for the region n >:: ; A::l
ne, i.e.
ne
>:::!A L
nnode
8Wint 8ujF~n\ue) = 8uTFint(u), (123)

e=l I=l
with F~nt(ue) = J Bj(pl + 'fh)dv, (124)

Qe
where Ue contains the components of the displacement vector associated with element e. The
constant pressure j5 is given explicitly by (118)2, while the matrices ([ h, B I and 8ui have the
forms given in (122), (121 )2 and (119) 2 , respectively. Note that the abbreviation F~nt defines the
internal nodal forces, while the global column matrices Fint and 8u define the internal forces
and the virtual displacements for all nodes of the finite element mesh, respectively.
By means of the expressions (113) 1 and (114h we find from the approximation (110) the
external virtual work ow ext' i.e.
ne
>:::;-A 2:= 8u'JFrt- A 2:= 8ujF<;,."j = 8uTFext,
nnode nb bnode
owext (125)
e=l I=l b=l I=l
with Frt = J N1bdv, F~xj = J

an,.b
N1tds, (126)
fle
where the column matrices b and t contain components of the body force and the surface trac-
tion, respectively. In eq. (125), the abbreviations F~xt and F~xj define external nodal forces,
associated with b and t, while the global column matrix pext captures the loads which act on the
continuum body occupying the region n with boundary surface an.
By means of the expression ( 115)1 we find finally from the approximation ( 111) the kinetic
virtual work <>wkin for the region n in the form
ne
L L
nnode nnode
8Wkin ~A 8ujMJKUK = 8uTMu, (127)

e=l J=l K=l
with MIK =I j N1pNKdv, (128)

fle
where UK and 8u1 are given by (119), and M IK defines the consistent mass matrix related
to the element domain De. The global mass matrix M and the global column matrix u, which
consists of accelerations, are calculated as an assembly of contributions arising from all finite
elements ne in the mesh. Note that, in general, the mass matrix will not be diagonal; however,
for practical purposes M is often diagonalized, which leads to the lumped mass matrix (see, for
example, Hinton et al. (1976) and Hughes (2000)). Note the equivalence of the tensor equations
(112)-(115) and the matrix equations (123)-(128).
With the definitions of the internal and external nodal forces and the mass matrix (see the
expressions (124), (126), (128)), and by means of the virtual work expressions (123)2, (125) 2
and (127)2 we obtain the associated form of the modified virtual work equation (94). Thus,
JuT R(u, t) = 0, R(u, t) = Mu(t) + pint(u(t))- pext(t), (129)
where residual forces form the entries of the column matrix R, which is referred to briefly as
a residual. It is important to note that some nodal displacements of the finite element mesh are
known. Hence, for nodes I lying on the discrete boundary surface anu
c {)[!the vector UJ is
given, while the corresponding values 8u1 at these nodes are zero.
Since eq. (129) must hold for all vectors 8u, we obtain
R(u,t) = 0, M u(t) + pint(u(t)) = pext (t), (130)
which are the semidiscrete equations of motion (discrete in space and continuous in time). Here
pint is a nonlinear function of the displacement u. Equations (130) form a system of N non-
linear ordinary differential equations in the nodal displacements, where N is the number of
unprescribed degrees of freedom. Those equations which correspond to the prescribed degrees
of freedom are ignored. The second-order differential equations are supplemented by initial con-
ditions for the given displacement and velocity fields at the instant of timet= 0, i.e.
uit=O = uo, (131)

148 G.A. Holzapfel
The matrix equations (130) and (131) define the discrete problem for fully nonlinear elastody-
namics.
If the data depend on time and the contributions due to dynamical quantities are negligible,
we have a quasi-static problem and (130)z takes on the form
~int(tt(t)) == ~ext(t), (132)
with the initial condition tt It=O == tto. Hence, at each time t we seek a solution tt( t) that makes
the internal forces equal external forces. In contrast to the dynamic (transient) problem, the vari-
able t in eq. (132) may not correspond to the actual time. This is the case, for example, when
rate-independent problems are considered. Finally, note that if the data are independent of time
the body is considered to be in static equilibrium and the set of equations reduces to the associ-
ated nonlinear boundary-value problem of elastostatics.
4.3 Consistent Linearization

In order to obtain solutions of the nonlinear initial boundary-value problem, as introduced above,
incremental/iterative solution techniques of Newton's type are often applied to solve a sequence
oflinearized problems. This strategy implies a consistent linearization of the nonlinear equations
with respect to the positions of the nodal points, which yields the corresponding geometrical
and material stiffness matrices. The notion 'consistent linearization' means a linearization of
all quantities associated with the nonlinear problem. The procedure of linearization is a very
important task and a key element in computational mechanics. The underlying technique was
first introduced in the mechanics of solids and structures by Hughes and Pister (1978).
We recall that we started our discussion with the Simo-Taylor-Pister variational principle,
which treats the displacement, pressure and kinematic field variables (u, p, ]) as independent
variables. In the next two steps we derived the necessary stationary conditions for the function
LsTP and provided the associated discretizations. Based on the chosen discretization it turned
out, however, that t~e volumetric variable~ are now dependent variables. In particular, the pres-
sure j5 depends on J via eq. (118)2 and J depends on the deformation via eq. (117)2. Hence,
both variables are subject to linearization, whose implementation is an important component.
Consequently, due to the interpolation functions used it is crucial to start with the spatial dis-
cretization of the Euler-Lagrange equations in the weak form and to continue with the lineariza-
tion of the discrete equations with respect to the positions of the nodal points. To linearize the
Euler-Lagrange equations in the weak form first and then to discretize them would not be appro-
priate; the operations of spatial discretization and linearization do not commute in the theory of
interest here.
Linearization in the Lagrangian description. In the following the linearization procedure

is presented at the element level. Since we have considered the loads as independent of the
deformation of the continuum body, the linearization only affects the modified internal virtual
work 8W~nt. Rather than linearizing the discrete relation (112)3, which is expressed in terms
of spatial quantities, we perform the linearization of its material version and then do a push-
forward operation on the material tensors. By recalling (95)2 we have the expression 8W~nt ==
fnHQe (JhphCh 1 + Sh) : 8Ch/2dV in terms of material quantities for a typical finite element, on
which we will focus subsequently.
It is important to emphasize that variation and linearization of functions are equivalent op-
erations, which are based on the concept of directional derivative. Hence, in eq. (82), we just
use the linearization operator .:1( •) instead of £5( • ), and have the similar relation L\4i(x, L\u) =
DLluP(x) = ddc:P(x + c.du)lc:=O· For a detailed description of the concept of linearization the
reader is referred to Holzapfel (2000), Section 8.4. By adopting rule (82) and by using the product
rule, we arrive at the form
(133)
of the linearization, where D Llub"C characterizes the directional derivative of b"C in the direc-
tion of the increment L\u of the displacement field. Note that the domain of integration in (133)
remains fixed (it does not depend on the deformation), which is particularly useful for the lin-
earization procedure. For convenience, we omit subsequently the subscript ( • )h indicating the
discretization.
We may then proceed to specify the directional derivatives of the two quantities JpC- 1 and
S. We use the chain rule to obtain
1
DLlu(JpC- +S)=
- (
2p
a(Jc- 1 )
ac +2ac
as) 1
:2DLluC+DLluPJC
-1
= Jp(C- 1 0 c- 1 - 2C- 1 8C- 1 ): ~DLluC

- 1 -1
+C: 2DLluC + DLluPJC , (134)
where the tensor product -C- 1 8 c- 1 is given by eq. (65). We introduced the definition C =
2(8S/8C) of the fourth-order elasticity tensor, which depends upon the material used. The ex-
plicit expression for C has the form
c = 2 ac
as = · c ··
:r · :rT + ~Tr(J-
3
2 1 3 s*)JP- ~(c- 1
3
0 s + s 0 c- 1)
'
(135)
with the abbreviation
(136)
for the fictitious elasticity tensor C* in the material description. The modified projection tensor
lP and the trace Tr(•) are defined by (64). For a detailed account ofthese results the reader is
referred to Holzapfel (2000), Example 6.8. A constitutive particularization of eqs. (135)2 and
(136)2 for arterial wall mechanics is presented in Section 3.3.
By means of (134)2, the linearized relation (133) may be re-expressed as
-· I
DLlub"W~nt = 1
{(JpC- 1 + -S): 2DLlub"C 1
+ DLluPJC- 1 : 28C
Doe
150 G.A. Holzapfel
For notational simplicity we now use L1p for D LluP and L1C for D LluC. In addition, we use
the explicit expression D Llu8C/2 = sym(GradT L1u Grad8u) for the first term of the integrand
in eq. (137) (see Holzapfel (2000)). Herein, Grad(•) (with uppercase G), denotes the material
gradient of the material field ( • ), which is obtained by differentiating (•) with respect to the
referential position X, at a fixed timet. By taking advantage of symmetries associated with C and
S the linearization of D Llu8W~nt in the material description leads to the set oflinear increments
+ ~8c: [(c- 1 0 c- 1 - 2C- 1 8 c- 1 )Jp + CJ : ~L1C}dV. (138)
Linearization in the Eulerian description. The next goal is tore-express the result (138)
in the spatial description. This is done by means of appropriate push-forward operations on the
material tensors. The details, which can be found by Holzapfel (2000), Section 8.4, are left to be
supplied by the reader. Here we summarize the result for the linearization of D Llu8W~nt in the
spatial description, i.e.
D Llu8W~nt = J {grad8uh : gradL1uh (phi+ uh) + L1ph I : 8eh

De
(139)
where 8eh is given by (107), and the discretization

l
= sym(gradL1uh) = 2 L
nnode
L1eh (L1ur 0 VxNr + VxNr 0 L1ur) (140)

I=1
of the increment of the Euler-Almansi strain tensor is the analogue of eq. ( 107). The superscript
( • )h in (139) is to be used again to indicate discrete quantities. We introduced the notation c for
the elasticity tensor in the spatial description, which is defined as the push-forward operation of
C times a factor of J- 1 . It is the Piola transformation ofC on each large index so that
(141)
Equations (138) and (139) have a similar symmetric structure. They are linear with respect
to the terms 8u and L1u, depending on X and x, respectively, and lead to a symmetric (tangent)
stiffness matrix. Formulations according to (138) and (139) are sometimes in the literature called
total-Lagrangian and updated-Lagrangian, respectively. This really means that integrals are cal-
culated over the respective regions of the reference and current configurations. It is important to
emphasize that the material representation (138) of the linearized modified virtual internal work
is equivalent to the spatial version ( 139). The two representations are based on the use of change
of variables, and the results are the same in both cases.
For convenience, we may write the linearization ofthe discrete modified internal virtual work
(139) in the decoupled form
(142)
with the definitions
(143)
(144)
D Llu8We~>TP = J .dph I: 8ehdv (145)

fle
of the three contributions. The first term in eq. (142) comes from the current state of stress
and represents the geometrical stress contribution to the linearization, while the second term
represents the material contribution. The last contribution D Llu8W~TP governs the change of
Ph and is due to the mean dilation approach, which is based on the three-field variational principle
ofSimo, Taylor and Pister.
Derivations of the stiffness matrices. The aim now is to compute explicitly the three contri-
butions (143)-(145) and to provide the associated (tangent) stiffness matrices. Let us start with
the geometrical stress contribution D Llu8wro to the linearization, as given in (143), and study
the explicit discrete form for a typical element fle. By means of the interpolations (104) and
(108) 1 we obtain from (143) that
nfenfe J(8uJ&JVxNI): (.duK&JVxNK)(pl+uh)dv. (146)

1=1 K=1n"
For later use it is now beneficial to rearrange the integrand of this equation. The straightfor-
ward rearrangement is shown by means of the analysis of the term (a@ b) : (c@ d)A, which has
the same structure as the integrand in eq. (146) 2 , where a, b, c, dare vectors and A is a symmetric
second-order tensor. Basic tensor algebra gives
(a&Jb): (c&Jd)A= (c&Jd)T(a&Jb) :A

=(d&Jc)(a&Jb):A
=(a·c)(d&Jb):A
= (a· c)(b ·Ad)
= a. I(b · Ad)c, (147)
(see Holzapfel (2000)). Hence, applying the property (147) 5 to (146)2 we find the simple format
(148)
152 G.A. Holzapfel
This equation reads in the convenient matrix notation:
L L
nnode nnode
DLlu8wro = 8ujKf;LluK (149)

1=1 K=1
with KJe; =I jCVxN1)TCfo'Y'xNKdv, (150)

Qe
where KJe; is called the geometrical stiffness matrix, which defines the stress contribution to
the tangent matrix. It arises from the finite deformation kinematics of the considered problem
and relates nodes I to K within a typical finite element domain. The sub-stiffness matrix KJe;
has the size nctof x nctof, where nctof denotes the number of degree of freedom associated with
the element node (for a 3D continuum problem we have nctof = 3 for each node).
The initial (Cauchy) stress matrix (containing stresses at the beginning of each iteration-step)
is indicated by ([ 0 . It is a square matrix with the structure
(151)
where the entries aij are found from the respective constitutive equation to be used.
Let us continue with the constitutive contribution D Llu8Wemat to the linearization. By means
ofeqs. (107)2, (108)1 and (140)2, we find from (144) that
This equation reads in matrix notation:
L L
nnode nnode
DLlu8Wemat = 8u'JK'FJ(tlluK, (154)
J
1=1 K=1
with K'F](t = BjDBKdv, (155)

f2e
where K'F](t is the material stiffness matrix for the element nodes I and K of a typical element.
It defines the constitutive contribution to the tangent matrix and arises from the resistance of the
hyperelastic material to imposed deformation. The (spatial) constitutive matrix is indicated by
D. The entries of D are the components of the fourth-order tensor (I Q9 I - 2JI)p + iCh. Taking
advantage of the minor and major symmetries of iCh we deduce that D = Dfi + D-e, with the
explicit expressions for the matrices
-p p p
p -p p
p p -p
Dp= (156)
-p
-p
-p
cuu cu22 C1133 cu12 cu23 cu13
c2222 c2233 c2212 c2223 c2213
C3333 c3312 c3323 C3313
D-e= (157)
c1212 c1223 c1213
sym. c2323 c2313
c1313
Let us finally specify the pressure part D Llu<~"W~TP, which is a crucial contribution to the lin-
earized quantity D Llu<~"W~nt. For this specification we require the interpolation of the increment
of the volumetric variables. Hence, by analogy with (108)2 and (108) 4 we write
(158)
By means of(107)1, the property I: sym(gradb"uh) = divb"uh, and (158)1 we find from (145)
that
where, for convenience, the definition
-
div(•) = - 1/
Ve
div(e)dv (160)
of the average divergence over the current volume Ve of an element is to be introduced.

Before examining the pressure ~hange lJ.p occurring in (159)3 it is first necessary to deter-
mine the directional derivative of J in the direction of the increment iJ.u of the displa~ement
field. Therefore, by means of (158) 2 , the linearization of (117) leads to LJ.]
= (1/Ve) fn"0 e iJ.JhdV. With the incremental property iJ.Jh = JhdiviJ.uh we find that the
change of the volumetric variable is given by
LJ.] = ~ J
f2e
diviJ.uhdv = J diviJ.uh, (161)
where the operator div( • ), as defined in eq. (160), is to be used.

154 G.A. Holzapfel
By recalling relation (118)2 and using (161 )2, we find an explicit form of the pressure change,
namely
(162)
Substituting this result into (159)3 gives
(163)
where J = vefVe, as shown in (117)2, and, for notational simplicity,"' h_as been introduced as
an abbreviation. It is important to emphasize that the variables .dp and L1J for the pressure and
volumetric quantities are eliminated at the element level.
The aim is now to express eq. (163)1 in the same mathematical structure as (148). By using
the definition (160) and the interpolation (1 05), a straightforward algebraic manipulation gives
L L
nnode nnode
="' 6u1 · (VxNJ l8l VxNK)L1uK, (165)

1=1 K=1
where the definition
-Vx(•) = -
1
Ve
f Vx(•)dv (166)
of the spatial average gradient of the scalar function ( •) over the current volume Ve of an element
is to be used.
The pressure part (165)2, which is given in tensorial form, is recast in the matrix notation as
L L
nnode nnode
D.1u6W;'TP = 6ujK7'Jl L1uK, (167)

1=1 K=1
with (168)
where K7jl is the 'pressure' stiffness matrix, which defines the important pressure contribution
to the tangent matrix.
The total stiffness matrix for a typical element, denoted by K e, is composed of the three sub-
stiffness matrices given through eqs. (150), (155) and (168), which is summed over the nodes of
the elements. Thus,
(169)
The size of the element stiffness matrix K e is then ( nnode · nctof) x (nnode · nctof), which, for an
8-node brick element, is (8 · 3) x (8 · 3) = 24 x 24.
Note that motion-dependent loads of a body cause additional external virtual work (for pres-
sure boundary loading see, for example, Holzapfel (2000), p. 383). The external virtual work
done by the loads along the virtual displacement need to be considered within nonlinear finite
element analyses in the form of an additional, in general non-symmetric, contribution to the ele-
ment stiffness matrix Ke (Schweizerhofand Ramm (1984)).
The construction of the global (tangent) stiffness matrix, denoted by K, follows the standard
assembly procedure of element stiffness matrices. We write
=A Ke(ue),
ne
K(u) (170)
e==l
with K e given in (169). Since both volumetric variables J and p are eliminated at the element
level we now have a (generalized) displacement model with only the global variable u unknown.
Hence, the implementation of the global stiffness matrix K, which is symmetric for hyperelastic-
ity, and the residual R (compare with (129)2) is the same as for the standard displacement-based
method in which only the displacement field is discretized. It is known that the stiffness matrix
K is the Jacobian matrix of the internal nodal forces according to
8Fint(u)
K(u) = au . (171)
Note that the derived Ql/ PO-element does not satisfy the Babu§ka-Brezzi, or LBB, stability
condition within the geometrical linear theory and instabilities may occur (Brezzi and Fortin
(1991)). For mathematical details of the LBB-condition the reader is referred to the book by
Oden and Carey (1984).
4.4 Solution Methods of the Spatially Discrete Equations
So far we have considered the discretization of the continuous problem in space. To complete our
study of finite deformation elastodynamics, we will apply a finite difference scheme to achieve
also discretization in time.
The task is now to find the displacement u( t) satisfying the nonlinear initial boundary-value
problem in the form of the spatially discrete equations (129)2 and the given initial data (130) over
some closed time interval t E [0, T] of interest, where 0 = t 0 < ... < tM+l = T. Subsequently,
we consider the time discretization (partition)
M
[0, T] = U[tn, tn+l], (172)
n==O
where n denotes an index on the closed time sub-interval [tn, tn+l], and M + 1 is the number of
the sub-intervals. We now concentrate attention on a typical sub-interval, with
(173)
156 G.A. Holzapfel
characterizing the associated time increment. For tracing a nonlinear elastodynamic problem not
all time sub-intervals need to have the same width.
Assume now that up to a certain time tn the algorithmic approximations for the solution Un,
the velocity Un, and the acceleration Un are specified uniquely by the given motion attn. The
aim is now to advance the solution to time tn+ 1 = Llt + tn and to update these quantities to
Un+1· Un+1, and Un+1·
In order to solve (129)2 and (130) for a typical time sub-interval we may use the most-widely
used (global) time-stepping algorithm by Newmark (1959); see also Hughes (2000), Chapter 9.
This algorithm consists of the equations
Miin+l + Fint(un+d = Fext(tn+l), }
Un+1 = Un + Lltun + Llt [(1 - 2j3)iin + 2/3iinH], (174)
itn+1 = itn + L1t[(1- !')iin + /'UnH],

where j3 and I' are (free) algorithmic parameters. Different specific choices of these parame-
ters result into different time-stepping algorithms. For comparative studies see, for example,
Goudreau and Taylor (1972). The Newmark method is second-order accurate - the truncation
error is of order ( Llt) 2 • The set of nonlinear algebraic equations ( 174) is solved in each time
increment. An iterative solution strategy is based on the efficient predictor/multicorrector algo-
rithms described in Hughes et al. (1979).
Other transient time-stepping algorithms such as the a-method by Hilber et al. (1977) may
be considered. The method controls (damps) the amount of high frequency dissipation, while
the predictor and corrector formulas ofNewmark's method are retained. A variant of the method
is found in Chung and Hulbert (1993). For another modification of Newmark's method see, for
example, Wood et al. (1981 ).
Explicit methods. An explicit finite element integrator is the central-difference method widely
used in structural dynamics. The algorithm is deduced from the Newmark formulas (174) by
using the central difference as a template, i.e. j3 = 0 and I' = 1/2. It has the form (Wood
(1990))
..
Un+1 = M-1 [Fext(t n+1 ) - Fint(Un+1 )] ,
;\ . L1t2 ..
Un+1 = Un + LltUn + - 2-un, (175)
.
Un+1 = Un
. +2
Llt (.. + .. )
Un Un+1 ·
As can be seen from the decoupled eq. (175)2 the displacements at time tn+1 can be computed
explicitly from the three matrices {Un, Un, Un}, which are known at tn. Since Fext at tn+ 1
is given, accelerations at tn+l follow from the time discrete equation of motion (175h, and,
finally, the velocities may be computed from eq. (175)3. Hence, the explicit method is simple to
implement since the required solutions at tn+ 1 only depend on the quantities attn. For the case
that the mass matrix M is diagonal (mass lumping), the inverse M- 1 of M is trivial to compute
and the integrator turns out to be particularly efficient.
The disadvantage of explicit finite element methods is the fact that they are only stable when
Llt is less than a critical value. Hence, explicit methods are particularly suited to compute struc-
tural responses in the high frequency domain, for which small time steps are often required to
capture the associated physical process. To resolve vibrations in the low frequency responses,
this method may require an inefficiently large number of time steps due to the stability limit.
Hence, for this type of problem explicit integrators are the methods of choice.
Implicit methods and aspects of nonlinear equation solving. One example of an implicit
finite element integrator is the trapezoidal rule (average acceleration method). This method is
widely used for structural dynamics applications and is deduced from ( 174) by taking (3 = 1/4
and')'= 1/2. Thus,
M Un+l
oo
+ pint(Un+l ) = pext(t n+1 ) ,
(176)
In practice, Un+l is eliminated from the time discrete equation of motion (176h by means
of ( 176)2. This results in a highly nonlinear and coupled algebraic equation for u n+ 1, given by
.1~ 2 Mun+l + pint(un+l) = pext(tn+l) + M [un + .::1~ 2 (un + Lltitn)]. (177)
By writing eq. ( 177) in the structure of ( 130) we obtain the nonlinear discrete equations of motion
(using the trapezoidal rule as a template) in the form
(178)
with
R( Un+l) = pext (tn+l) - pint (Un+l) - .::1~ 2 M Un+l
+M [un + .::1~ 2 (un + Lltun)], (179)
where R( Un+ 1 ) is the column matrix of the residual forces, and the unknowns u n+ 1 are a set of
nonlinear functions.
The task is now to find the unknown quantities at tn+ 1 , in particular the column matrix Un+ 1 ,
by solving the set ( 178) of nonlinear algebraic equations using iterative solution techniques. A
wide variety of such solving strategies exist in computational mechanics.
One representative example of a solution method for nonlinear algebraic equations is the
Newton-Raphson method. Although this method is well-established and in common use, the
method is insufficiently robust and is associated with high computational cost. It is an equi-
librium iteration, which exhibits quadratic rates of convergence once we are close enough to
the solution point (local convergence), pre-supposing that 'correct' linearization of the nonlin-
ear equations is provided. There are many other iterative solution schemes available, such as the
modified Newton method, which shows only linear rates of convergence.
158 G.A. Holzapfel
Now, within the context of the present volume, we will review briefly the application of the
Newton-Raphson solution algorithm to the implicit finite element method. The linearization of
(178) may be written as
(180)
where u~+l is the displacement at iteration i and u~~\ = u~+ 1 + .du is the new displacement
at iteration i + 1. Note that the last term in the linear equation ( 180) is the directional derivative
of R with respect to the displacement estimate u~+l in the general direction of the increment
.du. Hence, with the discrete equation (180)2, the general Newton-Raphson iteration scheme
becomes
(181)
which is to be solved for an increment .du in the solution followed by the displacement update
(181)2. Here we iterate within each time step until, for example, the magnitude IIR(u~+ 1 )11 of
the residual (i.e. the Euclidean norm calculated from the current iterated value), is less than a
pre-assigned tolerance that ensures an accurate approximate solution to the nonlinear algebraic
problem. One simple convergence test is based on the comparison of IIR( u~+ 1 ) II with a fraction
of the initial residual IIR(u~+ 1 )11 such that IIR(u~H)II < EIIR(u~H)II is satisfied, where
E is a pre-assigned, non-dimensional tolerance parameter. Another measure of accuracy is, for
example, the energy norm determined by the expression IIL1u T R( u~+ 1 )11 (for more details see,
for example, Belytschko et al. (2000) ). If the convergence test is satisfied the time sub-interval is
updated and the Newton-Raphson iteration starts with the next step to advance the solution.
By using this concept and by taking u~+l instead of Un+ 1 in relation (179), we obtain from
(181 )I the linear set of equations for the displacement increment .du, i.e. the iterative driver
(182)
where the residual at iteration i is given as
R( un+1
i ) pext (t ) pint ( i ) 4 M i
= n+1 - un+1 - .1t2 un+1
(183)
and the stiffness matrix evaluated at the current value u~+ 1, i.e. at each iteration i and time tn+1,
is of the form
(184)
The stiffness matrix (184) is the time discrete version of (171 ). The displacement is then up-
dated via (181 )z. Note that the deviation of the approximate solution from the exact solution
decreases quadratically at each iteration when the linearization ( 184) is used. In practice, the
notion 'Newton-Raphson method' often refers to an algorithm in which K is formed at each it-
eration. So, for large-scale computations, the solution of the nonlinear algebraic problem may be
associated with high computational effort due to the formulation and factorization of K at every
iteration. Alternatively, modified Newton methods may be applied in which K is computed only
once (at the first iteration) and then kept fixed during the remaining iterations ofthe time step, or
even the entire solution strategy. This widely used method requires more iterations to converge
and does not show the superior convergence properties of the Newton-Raphson method, but the
computational cost per iteration is less. It is used mainly for weakly nonlinear problems.
For the quasi-static case the inertia terms are negligible, and hence eqs. (182) and (183) may
be reduced to the specific forms
R( un+l
i )
= pext(t n+l ) - pint( un+l
i )
' (185)
where the stiffness matrix K(uh+ 1 ) is given through eq. (184). The Newton-Raphson iteration
boils down to the solution of the sequence of linearized problems (185), so that equilibrium is
restored at tn+I, which results in satisfaction of (178).
The advantage of implicit finite element methods over explicit methods is the fact that for lin-
ear transient problems and suitable implicit integrators the time steps Llt can be chosen arbitrarily
in order to obtain unconditionally stable solutions. However, for large time steps the accuracy of
the solution decreases. Although no general proof of unconditional stability for nonlinear sys-
tems exists, we are taught by computational experience that time steps within implicit methods
can be chosen much larger than within explicit methods while retaining stability. Hence, im-
plicit integrators are particularly suited for computing structural responses in the low frequency
domain, for which larger time steps are desirable.
Line search method. The Newton-Raphson method converges slowly if the starting iteration
is substantially far from the (exact) solution (Luenberger (1984)), so larger time steps reduce
the robustness of the Newton-Raphson method. In addition, the modified Newton-method may
also lead to poor convergence rates due to the fact that the affiliated stiffness matrix is only a
rough approximation to the consistent one, as evaluated in (184). Consequently, the displace-
ment updates may not be effective any more in the sense that the updated values are further away
from the desired solution. Since Newton-type methods have an unfortunate tendency to wander
off if the initial guess is not sufficiently close to the root, it is therefore necessary to enhance
these methods by 'post-processing' strategies to preclude divergence and to improve the conver-
gence rate. The enhancements should be performed in an economical way. The widely used line
search method, discussed briefly here, has proven to be effective in controlling the solution and
preventing divergence of the nonlinear iteration through Newton-type methods.
Basically, the line search method interprets the increment Llu, found by the Newton method,
as a 'search direction'. It is scaled by a (scalar) search parameter, denoted by s. Rather than
using the update formula (181 )2, the new displacement is now computed by the formula
i+l
un+I (s) = un+I
i
+ sLlu, (186)
where u~~\ and uhH denote the new and old solutions. For nonlinear dynamic (transient) prob-
lems the direction .1u is determined by solving (182), with the use of eqs. (183) and (184), while
160 G.A. Holzapfel
for nonlinear quasi-static problems Llu is given through eqs. (185) and (184). The aim is now
to find the search parameter s such that some norm on the residual matrix is minimized with
respect to the non-scaled (standard) update, i.e. the full Newton step s = 1. Note that the line
search method becomes less effective in improving the convergence rate of Newton-Raphson
iterations the closer u approaches the solution point.
Now we define a scalar-valued nonlinear function g( s) such that the search parameter s
satisfies the condition
(187)
s
This scalar equation for admits a general representation of the residual Revaluated at u~~\ =
u~+l + sLlu. For nonlinear dynamic problems the residual is given through eqs. (183) and (186)
using the trapezoidal rule as a template, while for quasi-static problems the residual is given by
(185). The task is now to design a line search algorithm that selects s in such a way that at the
end of each iteration the search direction Llu is orthogonal to the matrix R( u~~\) forming the
residual forces.
The condition for s in the form of the definition (187) has to be satisfied to some tolerance,
i.e. g( s) ~ 0. There is no explicit expression for the search parameter s in the presence of
nonlinear problems. Hence, a solution procedure is required for solving the nonlinear scalar
problem g( s) = 0 for the single scalar unknowns. One possibility is to use again a Newton-type
method, however, the solution procedure aims to be efficient so as to keep the computational costs
of the line search algorithm as low as possible. In practice, the condition g( s) = 0 is regarded as
too stringent and it would not be efficient (or not even necessary) to apply an 'exact' line search.
One accepts convergence of the line search process with a relatively large pre-assigned tolerance.
A line search algorithm may return values of s > 1, which in practice may lead to poor
convergence rates for nonlinear problems. Hence, the root should be expected in the interval
between 0 and 1. By substituting s = 0 and s = 1 into (187) we may deduce the condition
g(1) · g(O) < 0. If this condition is satisfied then sis determined by some sort of root finder. The
search parameter s is then accepted if the condition
lg(s)l ::; E lg(O)I, (188)
holds, where E is a user-specified (line-search) tolerance parameter. A value of E = 0.5 is pro-

posed by Matthies and Strang (1979). For alternative convergence criteria the reader is referred to
the books by Luenberger (1984) and Bazaraa et al. (1993). A widely used line search algorithm
is based on residual measures that are interpolated by a quadratic function of s (see, for example,
Bonet and Wood (1997)).
Another enhancement for Newton-type methods is the quasi-Newton method (also called the
secant method) with the BFGS-update (named after its creators: BROYDEN, FLETCHER, GOLD-
FARB and SHANNO ). For more details see, for example, Matthies and Strang (1979), Dennis and
Schnabel (1983) and Luenberger (1984).
Finally, we wish to emphasize that Newton's methods are not appropriate for tracing nonlin-
ear responses near limit points (one example is the quasi-static computation of the snap buckling
phenomena of structures). The stiffness matrix approaches singularity in the neighborhood of
limit points and the solution diverges. So-called arc length methods are well-suited to constrain
the iterative solution. The description of this topic is beyond the scope of this article, the reader
may consult the survey article by Riks (1984). For more details of the implementation techniques
of solution algorithms within a finite element program for both dynamic and static nonlinear
problems the reader is referred to the books by Crisfield ( 1991 ), Crisfield (1997), Bathe ( 1996),
Zienkiewicz and Taylor (2000), Belytschko et al. (2000) and Wriggers (2001) among many oth-
ers.
5 Representative Numerical Examples

The following numerical examples are supposed to be representative in the sense that they aim
(i) to show the performance of the structural arterial models outlined in Sections 2 and 3, and (ii)
to document finite element results that are in good qualitative agreement with experimental data.
For this purpose we consider an arterial segment with no pathological changes of the innermost
layer. Hence, we approximate the arterial segment by two separate thick-walled fiber-reinforced
circular layers, i.e. the media and the adventitia, which behave incompressibly.
The first example is concerned with a finite element analysis of the mechanical behavior
of an artery during clamping. The three-dimensional deformation and the stress distributions
in the arterial wall during the clamping procedure are presented. The remaining two numerical
examples are concerned with the investigation of the characteristic viscoelastic (relaxation and
creep) behavior of a healthy young artery under various (static and dynamic) boundary loadings.
All three-dimensional nonlinear finite element analyses have been performed with the mean
dilatation Ql/ PO-element and on a HP-17000 workstation under the UNIX operating system. A
detailed account on the underlying three-field variational principle, the finite element discretiza-
tion and the consistent linearization procedure was provided in Sections 4.1-4.3. The proposed
structural models have been implemented in Version 7.3 of the multi-purpose finite element anal-
ysis program FEAP, originally developed by R.L. Taylor and documented by Taylor (2000).
5.1 Elastic Behavior of an Artery During Clamping

Arterial clamps are used to compress arteries during surgery so that blood flow is arrested. On
the basis of the elastic arterial model, as outlined in Section 2, a three-dimensional finite element
model for arterial clamping is reviewed (see Gasser et al. (2002) and references therein). The
numerical model provides a tool for studying the layer-specific mechanical response of clamped
arteries subject to an arbitrarily chosen clamp design.
In Section 2.4 the numerical example of the deformation behavior of a carotid artery during
combined bending, inflation, extension and torsion has demonstrated that it is essential to incor-
porate residual stresses (and strains) in the load-free configuration. In order to consider residual
stresses associated with the load-free configuration D0 , we introduce opened-up (reference) con-
figurations for the media and adventitia. Each arterial layer in the reference configuration is
assumed to be unstressed (the residual stresses are entirely removed by leaving all other proper-
ties of the material unchanged) and taken to correspond to an open sector of a circular cylindrical
tube with opening angle a, wall thickness H, inner radius Ri and length L, as indicated in Fig-
ure 8. We assume that the media occupies 2/3 of the arterial wall thickness and that the collagen
fibers are helically wound. Their orientations at a reference point X, characterized by the angle
cp, are different for each layer. Specific geometrical data for each arterial layer are summarized in
Figure 8. The arterial clamp is idealized as a pair of (rigid) cylinders with a radius of 3.0 (mm).
162 G .A. Holzapfel
RiM = 3.302 (mm)

.:::
'1:1 HM = 0.493 (mm)
:E
cu
<j)M = 10.0°
O"M = 160.0°
L r--;;-' R; A
\ :c = 3.482 (mm)
' ' .... :c
c: HA = 0.247 (mm)
' ' ...cu C{JA = 40.0°
"C
< O"A = 120.0°
Figure 8. Opened-up (stress-free) configuration of an arterial layer and associated geometrical data for the
media and adventitia.
The (elastic) material parameters Ci, k1 i, k2 i, i = M, A, which are involved in the strain-
energy functions (38) and (39), were fitted to the experimental data of a human left anterior
descending coronary artery (LAD), as given in Carmines et al. ( 1991 ). For purposes which were
made clear on p. 18 we set CM = lOcA. The penalty parameters r;,i, i = M, A, see eq. (80), were
chosen to be 104 (kPa). The resulting values are summarized in Table 1.
Media Adventitia
CM = 27.0 (kPa) CA = 2. 7(kPa)

k1 M = 0.64 (kPa) k1 A = 5.1 (kPa)
k2M = 3.54 (- ) k2A = 15.4 (-)
KM = 104 (kPa) "-A = 104 (kPa)
Table 1. Elastic material parameters c;, k 1 ;, k2 ;, i = M, A, for the media M and adventitia A, and penalty
parameters "-i.
Finite element model. Due to the symmetry of the considered geometry and the loading
condition of the problem, only a wedge of the artery with 90° is discretized by 1200 Q1/ PO-
elements - 750 for the media and 450 for the adventitia. We used eight finite elements across
the wall thickness with the same radial dimension within a layer (five for the media and three
for the adventitia). We discretized the circumferential and axial directions of one arterial layer
by 10 and 15 finite elements, respectively, while we considered a refined mesh around the clamp
region. The nodes at the media/adventitia interface are linked together. The number of elements
used turns out to be sufficient to achieve accurate numerical results. The common symmetrical
boundary conditions for the structure are applied. During the whole computation the top and
bottom faces of the artery are modeled as remaining planar.
In order to handle the contact problem (the interaction of the boundary of the clamp with the
boundary of the artery) we use a point to surface strategy. The cylindrical surface of the arterial
clamp plays the role of a slave surface modeled by 121 slave points, which are nodes. The master
surface is associated with the (outer) boundary of the adventitia, at which contact is expected. It
is modeled by master points which are interpolated from 80 four-node (contact) facets.
Finally, frictionless contact is assumed between the arterial boundary and the boundary of the
clamp. The control of the penetration between the clamp and the artery is performed by means
of the penalty method. The discrete (contact) problem is summarized in Table 2.
Arterial wall (wedge of90°) II Arterial clamp

Q1/ PO-elements
Media 750 I Slave nodes I 121
Adventitia 450
4-node (contact) facets Outer boundary of the 80
adventitia (master surface)
Table 2. Summary of the discrete contact problem (wall/clamp interaction).
Loading process. The loading process of the two-layer structure is separated into three steps:
I. The stress-free arterial segments, with geometrical data given in Figure 8 and parameters as
summarized in Table 1, are applied to an initial (pure) bending deformation (for more de-
tails see Section 2.2). This deformation process generates the unloaded but stressed circular
cylindrical shape of the arterial segment, for which the opening angle a = 0.0°. Then, the
nodes at the media/adventitia interface (i.e. the basal lamina extema) are linked together.
Finally, we end up with homogeneous stress and strain states in the circumferential and axial
directions, which we now use as the reference for subsequent deformation processes.
2. The circular cylindrical two-layer structure is then inflated up to an internal pressure Pi =
13.33 (kPa) (100 mmHg, which is the regular mean arterial pressure), and axially stretched
up to an assumed value of ..\z = 1.1. These loads are considered to be physiological, produc-
ing the physiological state of deformation.
3. The interaction of the boundary of the clamp with the boundary of the artery is finally com-
puted as a displacement-driven problem of the rigid cylinder (clamp). The arterial clamp is
placed perpendicular to the axis of the artery (for the geometrical set-up of the clamping
process see Figure 9(a)).
Deformed states of the clamped artery. The deformed states of the artery before and after
pinching off with a clamp are illustrated in Figure 9(a)(b) (the physiological configuration is
characterized by state@, while deformed state® shows the nearly occluded artery). In order
to clearly demonstrate the deformations the figure takes advantage of the symmetries by showing
one half of the artery. As can be seen from Figure 9(b), the region that is affected by the clamp-
ing mechanism is relatively small. Figure 9( c) illustrates different deformed states of the artery,
which occur during the clamping process. State@ represents again the physiological configura-
tion (before contact with the clamp), while states @-® show the decrease of the arterial lumen
due to clamping.
164 G.A. Holzapfel
(a) State@ (b) State®

Clamp modeled as
a rigid cylinder
Artery at the physio-
logical configuration
(c)
State@ State@ State@ State@ State@ State®
Figure 9. Deformed states of the artery (media and adventitia) at different displacements of the clamp. (a)
Artery at the physiological configuration before pinching off with a clamp. (b) Deformed state of
the artery, for which the maximum clamp displacement has been applied (full contact between
the clamp and the artery). (c) Different deformed states.
Evolution of stresses in the artery during clamping. Now we study the evolution of circum-
ferential and axial stresses at a representative point of the artery during clamping.
Three-dimensional finite element analyses showed that in the neighborhood of the clamp,
there is a tendency for circumferential (Cauchy) stresses in the wall to decrease and axial stresses
in the media to increase. To demonstrate the local evolution (change) of stresses in the artery
during clamping, we pick a representative point P, which is located at the inner surface of the
media as indicated in Figure 10.
In order to show the (local) stress evolution at P we introduce, for convenience, the two
(dimensionless) scalar quantities
ao az
S() = ---, Sz = ---, (189)
aophys a z phys
which determine the normalized stresses se in the circumferential direction and the normalized
stresses Sz in the axial direction. In eq. (189), ae and az denote the Cauchy stresses in the cir-
cumferential and axial directions, while ae phys and a z phys denote the associated stress quantities
in the physiological state@, at the same point P.
@ ® © ® ~ ®
4
II 4
.."'
"'"'
..
.."'...
t: 3 3
"'"'
-
"'
-;
".C
=
"'... "'
-;
~ ';::!
..
E u.
s, = - - <'I
.:... 2 U z phys 2 "Q
'Oj .~
~
-;
8...
"'
-;
ue
se= - - 0
8... ue phys z:
0 1
z:
0 0
0 0.5 1.0 1.5 2.0 2.5
Clamp displacement (mm)
Figure 10. Evolutions of the normalized circumferential and axial stresses (so, sz) with the clamp displace-
ment at a certain point P located at the inner surface of the media. Note the high increase of sz
with respect to the physiological configuration.
(a ) Media (b) Adventitia
Axial Cauchy
stress (kPa)
-5.00E+OO
3.00E+OO
1.10E+01
1.90E+01
2.70E+01
3.50E+01
Figure 11. Distribution of the axial Cauchy stress in (a) the media and (b) the adventitia at the final state of
the clamping process.
166 G.A. Holzapfel
In Figure 10 the (local) evolution of the normalized stresses se, Sz with the clamp displace-
ment are plotted. As can be seen, the axial stress s z increases significantly, while the circumfer-
ential stress se decreases with progressive clamp displacement. Remarkably, at the final state®
of the clampin..[process, with an occlusion of92.5% referred to the lumen diameter at the physi-
ological state (A) (the maximum dimensions of the lumen are 0.378 (mm) and 5.983 (mm)), the
value of the normalized axial stress s z at the considered point P is about four times higher than
that at the associated state @. Since the media is relatively vulnerable in regard to axial exten-
sion, this may be associated with medial damage and damage of the attached endothelial cells.
This is not an obvious effect, is not described in the literature, and cannot be investigated experi-
mentally. The relevance of stress analyses is underlined by the fact that even non-destructive load-
ing that exceeds the physiological limit leads to biological responses (Jackiewicz et al. (1996))
that might be clinically adverse.
Figure 11 illustrates the distribution of the axial Cauchy stress in the media and the adventitia.
As can be seen, the maximum values of the stresses in the media and adventitia are similar;
however, their locations are different.
Since the media is not designed to carry high axial loads (this is a task of the adventitia),
the clamping process may cause injury in the media. In view of the high circumferential stresses
at the inner surface of the media one may conclude that the intima is also highly deformed and
damaged (compare with the experimental observations of, for example, Slayback et al. (197 6)
and Margovsky et al. (1997); see also the work by Harvey and Gough ( 1981) grading the damage
of the arterial wall). It is important to note that the axial stress increase in the media due to
clamping is strongly affected by the shape of the clamp jaws.
For the distribution of circumferential stresses in the arterial layers at the load-free and phys-
iological configurations the reader is referred to Gasser et al. (2002).
5.2 Viscoelastic Behavior of an Artery under Various Boundary Loadings
Two representative numerical examples are investigated in order to demonstrate the reliability
and efficiency of the viscoelastic arterial model outlined in Section 3. On the basis of a healthy
and young arterial segment, which is under various (static and dynamic) boundary loadings, we
aim to point out some of the peculiar viscous effects of arterial deformations.
Elastic and viscoelastic material parameters. For both examples we take the elastic material
parameters for the media and adventitia, and the penalty parameters, to be as summarized in
Table 1 (for a description of these parameters seep. 53).
Based on the statements of Section 1.1 that viscous effects are attributed mainly to the me-
chanical behavior of smooth muscle cells, we associate the time-dependent response with the
media, which is the heterogeneous arterial layer containing a very large muscle content. For our
study we use experimental in vitro data for a human LAD (Gow and Hadfield (1979)), since a
human LAD has already been used for the identification of the elastic material parameters. As
mentioned in Section 3.2, the dissipation of arterial soft tissues in cyclic loading is relatively
insensitive over a wide frequency spectrum. In order to replicate this characteristic dissipative
response we choose a set of five Maxwell elements (m = 5) and a set of five relaxation times
T 1 , ... , Ts that cover a time domain of four decades. The free-energy factors (3 1 , ... , (3 5 are de-
termined in such a way that the dissipation is (approximately) constant between frequencies
f = 0.01 (s- 1 ) and f = 100.0 (s- 1 ). This frequency domain is sufficient for the investigation of
physiological loading. Note that a viscoelastic model based on a single relaxation time, such as
can be found in the literature, would never be able to describe the dissipation response of arterial
soft tissues under cyclic loading.
A detailed identification process for quantifying the viscoelastic material parameters involved
in the constitutive model, as outlined in Section 3, is presented in Holzapfel et al. (2002a), Sec-
tion 5.1. The complete set of viscoelastic material parameters for the media, which is taken for
both numerical examples, is summarized in Table 3.
Media
(3'{' = 0.353 (-)

7 1 = 0.001 (s)
(Jf' = 0.286 (-)

72 = 0.010 (s)
(3~ = 0.298 (-)

73 = 0.100 (s)
(3'{" = 0.285 (-)

74 = 1.000 (s)
(Jf' = 0.348 (-)
7s = 10.00 (s)
Table 3. Viscoelastic material parameters for the media.
Finite element model. The three-dimensional finite element analyses are based on 8-node
Q1/ PO-elements. Due to the axi-symmetry of the problems, only a sector of the structure (a
wedge of any angle) is discretized by eight finite elements (five for the media and three for
the adventitia) and analyzed. The nodes at the media/adventitia interface are linked together
and common symmetrical boundary conditions for the structure are applied. During the whole
computation the top and bottom faces of the artery are modeled as remaining planar.
(i) Axial relaxation and creep test. In the first example we compare the axial relaxation and
creep responses under combined bending, inflation and axial extension of the segment, with
(one-dimensional) experimental data of Tanaka and Fung (1974).
We consider the opened-up geometry of a circular cylindrical tube (Figure 8). Bending de-
formations are applied to the open medial and adventitial sectors to give a closed two-layer
thick-walled tube, which is then stretched to Az = 1.1 in the axial direction (simple tension)
and fixed subsequently at this elongation. These two processes are performed rapidly so that
no time remains for relaxation. Hence, we assume that the relaxation process starts at t = o+
(for computational reasons we have chosen the deformation period to be w- 7 (s)).
Now the axial reaction force, say Fz, is computed as a function of time and normalized
with the axial reaction force at t = o+. Figure 12 shows the relaxation of the normalized
axial reaction force, which is denoted by the ratio kp = Fz (A.z, t)/ Fz (A.z, o+), versus time
t at axial stretch ..\2 = 1.1. The (linear) decrease of kp is indicated by triangles, while
168 G.A. Holzapfel
T
1.0 ----- - 1.0
~
••
....
r..
\
/I\
..IC
c..> -<
~~
..IC
.s
=
·.c=
..
-=....'"'
!:
.... .."'
:..
'"'
o:l 0.9 I Creep at 0.9 "0
axial force
-;
..e
"0
~ Fz = 25.85 (mN)
=a
..
8 :z=
:z=
0.8 0.8
[ k>.. = 0.779 :-- ......

-·-i k p = 0.785 1
~--------r-------~--------~--------~
10 -~ 10-3 10-1
Timet (s)
Figure 12. Axial relaxation and creep responses of an arterial segment. Triangles show the relaxation of
the normalized axial reaction force kF = Fz (A.z, t)/ Fz (A.z, o+) versus timet at axial stretch
Az = 1.1. Squares show the creep of the normalized axial stretch k>.. = Az (Fz , t)/ Az (Fz, o+)
versus timet at axial force Fz = 25.85 (mN).
thermodynamic equilibrium (no change in the values of the state variables at any particle of
the system) is reached at k p = 0.785 after some time.
In order to compute the arterial response due to creep, we apply an axial force of Fz =
25.85 (mN) on the closed tube, which is then kept fixed during the creeping process. By
analogy with the above, the viscoelastic process starts at t = o+. The value of Fz is chosen
so that the axial stretch Az of the arterial segment is 1.1 at t = o+, which gives the same
initial conditions as in the relaxation test. Now the axial stretch Az is computed as a function
of time and normalized with Az = 1.1. Figure 12 shows the creep of the normalized axial
stretch, which is denoted by the ratio k>.. = Az (Fz, t)/ .Xz(Fz, o+) = Az/1.1, versus timet
at axial force Fz = 25.85 (mN). The (linear) decrease of k>.. is indicated by squares, while
thermodynamic equilibrium is reached at k>.. = 0. 779. In comparison with experimental data
(Tanaka and Fung (1974)) the results given in Figure 12 represent qualitatively the typical
viscoelastic behavior of arterial segments.
(ii) Dynamic inflation test. The second example investigates the dynamic inflation of the
arterial segment of the muscular type under different dynamical loading characteristics. The
16
It ... -+
15 h ... 100.0 (s-1 )
h ... 1.0 (s-1 )
J; ... 0.001 (s- ')
--
01
~
'-'
14
Ts .. . 0.0 (s- ')
~
...:::1
<II
13.33
"'"' 13
~
1:1.
-;
...=
...
-
<II
c 12
11
10.67 .
/4 h h
10
4.90 4.95 5.00 5.05
Inner diameter d; (mm)
Figure 13. Evolution of the inner diameter d; of an arterial segment subjected to static and dynamic internal
pressure p;. The solid lines show numerical results based on sinusoidal pressure loads with the
frequencies h =
100.0 (s- 1 ), /3 =
1.0 (s- 1 ) and / 4 =
0.01 (s- 1 ). The dotted curves show
numerical results based on the limiting cases h = 0.0 ( s- 1 ) and /5 -+ oo of the frequencies,
representing the stiffest and softest (elastic) responses, respectively.
inner diameter response is investigated in regard to changes in internal pressure and fre-
quency. Sufficiently slow and fast processes, for which viscous effects are absent, are studied
as limiting cases.
To achieve a closed two-layer thick-walled tube (pure) bending deformations are applied to
the open medial and adventitial sectors. The closed tube is then inflated up to the internal
pressure Pi = 13.33 (kPa), axially stretched up to Az = 1.1 (simple tension) and then fixed
at this elongation. We consider a time period of 106 (s) in order to compute the equilibrium
state of the arterial segment, which we use as the reference state.
Starting from this state we subject the arterial segment to a sinusoidal pressure load Pi (t)
varying between 10.67 (kPa) and 16.0 (kPa) and representing the change of blood pressure
between the diastolic and systolic phases. We study load cycles with the specific frequencies
of h = 100.0 (s- 1 ), h = 1.0 (s- 1 ) and !4 = 0.001 (s- 1 ). The evolution of the inner
170 G.A. Holzapfel
diameter di with internal pressure Pi for each of the given frequencies is plotted in Figure 13,
and indicated by solid lines. The hysteresis loops demonstrate the typical insensitivity to the
frequency over several decades.
As can be seen, the loading and unloading cycles involve about the same dissipation, which
is represented by the area between the curves (a measure of the non-recoverable energy). In
addition, the deformation responses are stiffer at higher frequencies, a fact which has already
been reported by several researchers (see, for example, Learoyd and Taylor (1966) and Fung
(1971)) and which could not be modeled using pseudo-elasticity. Note that the hysteresis is
quite small, which is due to the small amplitudes of the pressure loads.
We also investigate sufficiently slow processes, h = 0.0 (s- 1 ) (enough time remains for
the arterial segment to adjust itself internally), and sufficiently fast processes, f 5 ---+ oo. For
computational reasons we have chosen the frequencies to be h = w- 5 (s- 1 ) and f 5 =
105 (s- 1 ). These two limiting cases describe (reversible) processes during which the arterial
segment is in equilibrium at all times and are associated with stiffest and softest responses,
respectively. Viscous effects do not arise and the dissipation is zero. The numerical results
are plotted in Figure 13 and indicated as dotted curves.
6 Layer-specific FE-Modeling of Balloon Angioplasty using MRI and

Mechanical Testing
A detailed understanding of the mechanical procedure of balloon angioplasty requires three-
dimensional modeling and efficient numerical simulations. Following Holzapfel et al. (2002b ),
this section aims to expose a fully three-dimensional model for eight distinct arterial components
associated with specific mechanical responses.
The 3D geometrical model is based on in vitro MRl of a human stenotic post-mortem artery
and is represented by NURBS surfaces (Non-Uniform Rational B-Splines). The NURBS ap-
proach is proven to be efficient for representing complex geometric information processed by
computers. Mechanical tests of the corresponding vascular tissues provide the fundamental basis
for the formulation of large strain constitutive laws, which model the typical mechanical char-
acteristics of arterial components under supra-physiologicalloadings occurring during balloon
angioplasty.
The 3D finite element realization considers the balloon-artery interaction and accounts for
vessel-specific axial in situ pre-stretches. 3D stress states of the investigated artery during balloon
expansion are presented. The proposed approach provides a tool that has the potential (i) to
improve procedural protocols and the design of interventional instruments on a lesion-specific
basis, and (ii) to determine post-angioplasty mechanical environments, which may be correlated
with restenosis responses.
Simplified models, which do not consider either axial in situ pre-stretch, three-dimensional
geometries or material anisotropy, may lead to significant deviations of the stress states compared
with the reference model (see the comparative study in Holzapfel et al. (2002b)).
6.1 Introduction and Motivation

Balloon angioplasty and stenting are purely mechanical procedures, which aim to dilate stenotic
or occluded arteries in order to restore blood flow. A more detailed understanding of the un-
derlying mechanics of angioplasty and an optimization of the interventional protocols may lead
to significant improvements of the outcome. Surprisingly, only a small number of physical and
computational models concerned with balloon angioplasty have been published, although this
interventional treatment is of great and steadily growing medical, economical and scientific inter-
est (American Heart Association (2000)). Basically, a suitable model has to be based on realistic
experimental data, appropriate geometrical and physical modeling, and on efficient numerical
realization using the finite element method. In particular, the model must operate in both the
physiological and the supra-physiologicalloading domain.
Interventionists are interested in methods that allow determination of appropriate procedural
parameters such as diameter ratio (balloon to artery), balloon length, inflation pressure etc., as
well as optimal instruments such as stents with specific designs. Moreover, models that allow the
prediction of the primary success of interventional treatments on a lesion-specific basis may im-
prove patient selection, i.e. the selection of patients for which balloon angioplasty is the optimal
therapeutical choice. Additionally, for biomedical engineers, appropriate models may be helpful
in order to develop novel designs of interventional instruments such as balloons and stents and
to prove their performance on a computational basis. From a computational point of view these
types of problems are of particular interest since they involve a number of challenging fields in
computational mechanics, such as nonlinear continuum and contact mechanics and the numerical
simulation of nonlinear material behavior at large strains.
Existing models do not, unfortunately, consider essential mechanical characteristics of arte-
rial stenoses undergoing large deformations at therapeutic loadings, which are far beyond the
physiological domain (for a survey see Holzapfel et al. (2002b )).
6.2 Geometrical, Physical and Finite Element Modeling
MR image based geometrical modeling. An external iliac artery (male, 68 years) was excised
from a corpse during autopsy within 24 hrs after death. For subsequent mechanical testing only
fresh tissues were used. Axial in situ pre-stretch, defined as in situ length/ex situ length denoted
by .\ 8 , was 1.04. A straight segment of the artery (20.0 (mm)) with an eccentric stenosis was
scanned by means of high-resolution magnetic resonance imaging (cross-sections with an in-
plane resolution of 0.3 (mm) and an axial resolution of 1.0 (mm)). For each scanned cross-
section, the borders of the arterial components were traced manually by a series of points. The
traced cross-sections were combined along the arterial axis, which resulted in a set of 3D point
clouds representing the boundary surfaces of the components. Finally, each of the boundary
surfaces was represented as a NURBS spatial model (Hoschekand Lasser (1993) and Piegel and
Tiller (1997) ), which was fitted to the associated point cloud.
After scanning, the artery was cut through transversely into two halves. One half was used
for corresponding histological analyses to allow material characterization, while the other half
was dissected anatomically into its major components, which are non-diseased intima I-nos 1,
collagenous cap 1-fl (fibrotic part at the luminal border), fibrotic intima at the medial border
1-fm, calcification 1-c, lipid pool l-Ip, non-diseased media M-nos, diseased fibrotic media M-
f and adventitia A (Holzapfel et al. (2000b)). From these components, stripes with axial and
1 The abbreviation 'nos' is frequently used in histopathology and stands for not otherwise specified. In the
context of the present study it means 'no appreciable disease·, or, more precicely 'non-atherosclerotic'.
172 G.A. Holzapfel
Adventitia Media Intima Media

A non-diseased diseased
M-nos non-diseased Collagenous cap Lipid pool Calcification Fibrotic part M-f
I-nos 1-fl I-lp 1-c 1-fm
Figure 14. 3D reconstruction of eight distinct components of a human stenotic artery distinguished by two
major regions: the non-diseased wall (consists ofl-nos, M-nos and the associated part of A), and
the eccentric plaque region (consists ofl-fi, 1-lp, 1-c, 1-fm, M-f and the associated part of A).
circumferential orientations were cut out and underwent cyclic uniaxial extension tests with con-
tinuous recording of tensile force, stripe width and gauge length at a constant crosshead speed
of 1.0 mm/min. For most specimens preconditioning was achieved by executing five successive
loading cycles. Typical specimen geometries can be estimated from Figure 14. The calcification
and the lipid pool were excluded from the tests and considered as rigid body and incompress-
ible (very soft) solid, respectively. A description of the procedure and the experimental results
obtained can be found in Holzapfel eta!. (2000b).
Based on the classification scheme described above, a three-dimensional geometrical model
of the considered stenosis was generated using interpolation between the 21 serial high-resolution
magnetic resonance slices. Figure 14 shows the three-dimensional reconstruction of the eight dis-
tinct tissues of the human iliac artery. As seen from the figure, two major regions can be distin-
guished: the non-diseased wall which consists of the non-diseased intima 1-nos, the non-diseased
media M-nos and the associated part of the adventitia A, and the eccentric plaque region , which
consists of the collagenous cap 1-ft, the lipid pool 1-lp, the calcification 1-c, the fibrotic intima at
the medial border 1-fin, the diseased fibrotic media M-f and the associated part of the adventi-
tia A. Note that in contrast to young laboratory animals, the non-diseased intima 1-nos exhibits
considerable thickness (and mechanical strength). This is caused by intimal hyperplasia, which
is defined as proliferation of intimal cells and increase of extracellular matrix without any lipid
deposits. Arteries with diffuse intimal hyperplasia are non-diseased by definition (Stary et al.
( 1992)). For human arteries, thickness ratios of intima/media from about 0.1 to 1.0 or more are
documented (see the review article by Stary eta!. (1992)). Non-puplished measurements per-
formed in the authors' laboratory showed an average thickness ratio of intima/media/adventitia
of13 /56/31 for aged non-diseased iliac arteries Schulze-Bauer et al. (200 1).
Physical modeling. The resulting data sets, which contain geometrical information, histolog-
ical characterizations and mechanical behaviors serve as a fundamental basis for the generation
of the specific physical models.
The investigated arterial components are assumed to be (nearly) incompressible fiber-rein-
forced composites, which exhibit strongly nonlinear anisotropic responses and which undergo
large strains (Holzapfel (2001)). They show a pronounced stiffening behavior in the large strain
domain due to collagen fiber recruitment (see Holzapfel (200 1) and Humphrey (2002) ). Specific
parallel arrangements of the collagen fibers in form of fiber families (Rhodin (1980)) lead to
anisotropic material responses. For the arterial tissues investigated we observed that most tissue
types responded with abrupt failure when exposed to loads that are far beyond the physiological
range, while the non-diseased media responded with non-recoverable deformations before tis-
sue failure occured (for the stress-strain response of the non-diseased media, see Holzapfel et al.
(2000b )). The physiological range of tissue stresses is assumed to be in the order of 50-100 kPa
at the mean arterial pressure of 100 mmHg and axial in situ stretch. This is justified by exper-
imental data published by Cox (1976) for circumferential wall stresses in dog iliac arteries, by
Fung et al. (1979) for wall stresses of rabbit iliac arteries in both directions, and by unpub-
lished data of human aged iliac arteries performed in the authors' laboratory (Schulze-Bauer et
al. (2001)).
The non-recoverable deformations may come mainly from damage mechanisms and also
from plastic-type phenomena. In order to trace a practicable approach and due to the lack of
appropriate data about microstructural processes associated with non-recoverable deformations,
we decided to apply the theory of plasticity. For the associated continuum basis and algorithmic
formulation the reader is referred to Gasser and Holzapfel (2001). The calcification I-c and the
lipid pool I-lp are modeled as a rigid body and an incompressible solid (neo-Hookean material),
see eq. (35), with a very low (stress-like) material parameter c = 0.1 (kPa). All other arterial
components involved are based on the constitutive description as presented in Section 2.3. The
fiber orientations used for the constitutive models are determined by means of histology and
the image processing tools described in Section 1.2. It turned out that the arterial components
investigated exhibit a symmetric arrangement of the two fiber families.
Particular physical models of the arterial components were obtained by fitting the general
constitutive formulations to the corresponding experimental data using the Levenberg-Marquardt
algorithm. Best fit constitutive parameters characterizing the elastic and inelastic responses of
different arterial tissues, and the associated fiber direction angles used for the subsequent numer-
ical simulation are adopted from Holzapfel et al. (2000b) (see also Holzapfel et al. (2002b) ). All
material models involved have been implemented in the finite element program ABAQUS V5.8
(Hibbitt et al. (1998)).
Numerical modeling. Based on this particular approach the incompressibility constraint is en-
forced by means of an augmented Lagrangian method (Simo and Taylor (1991)), which provides
superior numerical performance compared with standard penalty methods (Holzapfel (2000)).
Finite element discretization. The use ofNURBS provides smooth surfaces and a suitable
basis for mesh adaption procedures that allow mesh refinement with respect to the (original)
reference geometry. Based on the geometrical model, a mesh generator creates finite element
meshes with a user-defined degree of refinement. For convenience, throughout the entire dis-
174 G.A. Holzapfel
10
Intima
Adventitia Media Media
A non-diseased non-diseased Collagenous cap Lipid pool Calcification Fibrotic part diseased
M-nos 1-nos l·fl l-Ip 1-c 1-fm M-f
Figure 15. Exploded view of the 3D discretization of a human stenotic artery.
crete structure, the numbers of finite elements in the radial, circumferential and axial directions
are constant. This allows easy access to the node numbers, as required for the prescription of
boundary conditions and contact surfaces. To capture the incompressibility constraint the robust
Ql/ PO-element, as described in Sections 4.1-4.3, has been used.
Figure 15 shows the mesh obtained in an exploded view, where the specific discretizations of
the individual arterial tissues can be seen. In the radial direction the intima, media and adventitia
consist of four elements each. The whole arterial model is generated with 3828 isoparametric 8-
node brick elements. Each finite element is associated with one out of eight distinguished material
responses. Thus, there is an abrupt change of material properties across element boundaries,
which separate different arterial components. Because of the lack of according experimental
data, no interpolation for a smooth transition of material properties was applied. Hence, stress
analysis seems to be more reliable within an arterial component than at its boundaries.
The (fully inflated) balloon is modeled as a rigid cylinder-shaped structure of hexahedral
finite elements with a diameter of 10.0 (mm) and a length chosen to be equal to the stretched
vessel segment. The assumption of a rigid cylinder is justified by the fact that fully inflated
angioplasty balloons behave as non-compliant cigar-shaped structures.
Boundary conditions and loading process. Regarding boundary conditions the artery is
fixed axially at one end, while the other end is exposed to displacements according to the axial
in situ pre-stretch. Additional boundary conditions are prescribed in order to avoid rigid-body
motions of the artery and the angioplasty balloon. During the experimental preparations a sig-
nificant axial retraction of the adventitia was documented after dissection from the remaining
arterial tissues (for a general discussion see Schulze-Bauer et al. (2002)). Therefore, we included
the observed adventitial axial pre-stretch in the numerical model, which was measured as 1.2.
Each of the discrete arterial components is tied to its adjacent components, i.e. we just prescribe
displacement continuities accross the boundaries of neighboring components. This is motivated
by the fact that the 'adherence materials' that interconnect the arterial components have small
thicknesses compared with the dimensions of the components (for the case of healthy arteries
these materials are called internal and external elastic membranes). Hence, even if these mate-
rials are highly deformable, only minor movements of the components against each other are
possible. Moreover, for diseased tissues these materials are rather stiff due to the high content of
collagen fibers. Cracking of these layers, which may occur during angioplasty procedures, is not
implemented in the proposed numerical model.
We consider three consecutive states of loading: (i) intraluminal (physiological) pressure of
13.3 kPa (100 mmHg, which is the regular mean arterial pressure) before balloon angioplasty,
(ii) full expansion of the angioplasty balloon (d = 10.0 mm diameter) in addition to the applied
intraluminal pressure, and (iii) intraluminal pressure of 13.3 kPa after balloon angioplasty. Note
that, during the whole angioplasty process a surface (pressure) load of 13.3 kPa is applied to the
inner boundary surface of the artery. Both the balloon inflation and the stent expansion are mod-
eled as displacement-driven processes. At the beginning of the unloading process the prescribed
displacements are then replaced by equivalent reaction forces at each node, which are driven to
zero.
Contact problem. For the modeling of the contact problem, which occurs between the
boundary of the arterial wall and the angioplasty balloon, we use the master-slave approach
(slave nodes are constrained not to penetrate the master surface). The contact constraint is applied
by means of the penalty method. Additionally, the dual contact problem (interchange master
and slave surface) is solved in order to avoid problems with the contact algorithms associated
with normal vector computations on the contact surfaces. Finally, frictionless contact is assumed
between the arterial boundary and the boundary of the angioplasty balloon.
6.3 Numerical Results and Clinical Implications
The aim of this section is to report three-dimensional stress results in the arterial components,
induced by balloon angioplasty, in the human stenotic artery investigated and to conclude with
clinical implications. All finite element analyses have been performed with the finite element
program ABAQUS V5.8 and on an UNIX-based HP-17000 workstation.
Numerical simulation of balloon angioplasty. Figure 16 illustrates exploded views of the

artery at full expansion of the angioplasty balloon (10.0 (mm) diameter) showing circumferen-
tial and axial Cauchy stresses in (kPa). Note that for the purpose of visualization a smoothing
algorithm was developed. Mean values of Gauss point stresses were computed for each element
in circumferential and axial directions. The stress value of a particular node is recovered by av-
eraging the mean values of all involved elements. Then the internal stresses are obtained by a
linear interpolation in the same manner as for the displacements. Generally, the global stress dis-
tribution is significantly higher (about 2 - 3 x) in the circumferential than in the axial direction,
as expected for the radial balloon expansion. In the non-diseased wall elastic tissue limits are ex-
ceeded, which causes non-recoverable deformations. They occur in the entire media M-nos and
in parts ofthe intima I-nos, indicated in Figure 16 by hatched regions. In the clinical context the
phenomenon of non-recoverable deformations, occurring predominantly in the media, is termed
'controlled vessel injury'. This is known to be an essential mechanism for the luminal gain of the
angioplasty procedure (Castaneda-Zuniga (1985)). For the non-diseased intima I-nos this may be
176 G.A. Holzapfel
800
----
"' 700
~
600
"'
"'
~ 500
"' 400
~
c 300
~
~ 200
E tOO
e
:::J
a 0
-tOO
300
200
"'c"'
c..
tOO
"'
~(I)
0
-;;
·;;:
< -tOO
-200
Adventitia Media Intima Media
A non-diseased diseased
M-nos non-diseased Collagenous cap Lipid pool Calcification Fibrotic part M-f
!-nos 1-fl 1-lp 1-c 1-fm
Figure 16. Exploded views of a human stenotic artery at full expansion of the angioplasty balloon
(10.0 (mm)) showing circumferential and axial Cauchy stresses.
interpreted as the occurence of intimal fissures (compare with section 'Methods'). Interestingly,
although the adventitia is modeled as a homogeneous material and has a rather regular geometry,
it exhibits a complex three-dimensional stress distribution, which strongly motivates the use of
a 3D model. The high-stress region of the adventitia at z = 0 (mm) is an effect of the rigid
calcification opposite to it and may also be caused by the prescribed boundary conditions.
In the non-diseased wall elastic tissue limits are exceeded, which causes non-recoverable
deformations. They occur in the entire media M-nos and in parts of the intima 1-nos, indicated
in Figure 16 by hatched regions. In the clinical context the phenomenon of non-recoverable de-
formations, occurring predominantly in the media, is termed 'controlled vessel injury'. This is
known to be an essential mechanism for the luminal gain of the angioplasty procedure (Castaneda-
Zuniga (1985)). For the non-diseased intima 1-nos this may be interpreted as the occurrence of
intimal fissures. Interestingly, although the adventitia is modeled as a homogeneous material
with a rather regular geometry, it exhibits a complex three-dimensional stress distribution, which
strongly motivates the use of a 3D model. The high-stress region of the adventitia at z = 0 (mm)
is an effect of the rigid calcification opposite to it and may also be caused by the prescribed
boundary conditions.
The (eccentric) plaque region shows a sandwich-like stress pattern in the circumferential di-
rection. Here the (liquid-like) lipid pool and the calcification are low-stress regions. They are
embraced by two stiff structures, i.e. the collagenous cap I-fl at the inner side, and the fibrotic
part of the intima I-fm and the diseased media M-fat the outer side. These components undergo
high stresses, which may lead to tissue failure. In particular, circumferential stress concentra-
tions(> 700 (kPa)) are seen at the center of the collagenous cap (indicated by the thick arrow in
Figure 16), which is a typical location for plaque rupture, as documented in pathological studies
(Richardson et al. (1989)). According to our experimental data the ultimate tensile stress of the
collagenous cap was not reached in the simulation performed. Rupture of the collagenous cap
is associated with the risk of subsequent thrombus formation in real-life procedures. In addi-
tion, tissue failure may also occur in the diseased media M-f, which may initiate layer dissection
(indicated by a thin arrow in Figure 16). This observation has not been recognized in previous
mechanical studies. Such an effect cannot be observed by means of standard clinical imaging
modalities. Nevertheless, in real-life interventions, it may cause, for example, intramural hem-
orrhage, which subsequently may lead to dissecting aneurysm formation. For this specific mor-
phology no stress concentrations are seen at the plaque shoulders, which are the lateral transition
zones between the atherosclerotic plaque and the non-diseased intima. However, tissue failure at
this location is a frequent clinical finding (The et al. (1992) and Mario et al. (1995)) and may
lead to dissection (delamination) of the plaque from the underlying arterial wall.
The proposed numerical model offers the opportunity to quantify the angioplasty-induced
changes of the mechanical environment on a component-specific basis. One quantity that char-
acterizes the changes in the mechanical environment is the local stress difference, i.e. post-
angioplasty stress minus pre-angioplasty stress. Figure 17 shows the circumferential stress dif-
ferences at z = 6.0 (mm) for the previous simulation, whereas the stress differences are plotted
onto the post-angioplasty configuration at 13.3 (kPa). Additionally, the luminal shape before
angioplasty at 13.3 (kPa) is indicated by a dashed curve in order to demonstrate the luminal
gain.
There are significant local and component-specific changes of the mechanical environment
indicated by a complex pattern of stress differences. In real-life interventions the stress differ-
ences that depend crucially on the interventional protocol used, may initiate component-specific
biological responses, which ultimately may lead to a renarrowing of the dilated artery. This patho-
logical process is called restenosis and is the major shortcoming of angioplasty (occurrence up to
40 % within six months - Fischman et al. ( 1994) and Macaya et al. ( 1996)). Arterial tissues may
be regarded as regulatory systems that control their physical and chemical environments. In other
words these tissues compare actual values of these quantities with set point values and react on
deviations with a specific biological response such as growth, remodeling, atrophy etc. in order
to restore the set point values. In our case the stress differences may be seen as deviations of
the actual stress state from the (tissue-specific) referential stress state. The subsequent biological
reaction may lead to restenosis and thus, cause therapeutic failure. The proposed model allows to
investigate the driving forces of this process, i.e. the stress differences, on a component-specific
basis. For the underlying mechanics of arterial remodeling the reader is referred to the works by
Rodriguez et al. (1994), Taber (1995), Section 7.2, Hayashi et al. (1996), Taber (1998), Rachev
et al. (2000), and references therein.
Special merits of the methodology. One aim of this section was to introduce and describe the
numerical methodologies of a model, which yields the most sophisticated simulations of inter-
ventional procedures to date. For this type of studies the proposed model will provide a suitable
computational basis in the future. Based on a large number of investigated stenoses the approach
178 G.A. Holzapfel
,....... 60
C<l
~
"-' so
0"'"' ,,
,
.b 40 '
"'>.
...<:::
I
I
I
<)
30 I
:::l
C<l ''
u I
\
20 \
:"§ \
c
\
0 - 10
'' '
..... ''
' '
~
E 0
:::l
<)
.....
u - 10
Figure 17. Differences in circumferential stresses (post-angioplasty stresses minus pre-angioplasty stresses
at 13.3 (kPa)) for a cross-section (z = 6.0 (mm)) after a full expansion of the angioplasty
balloon with a diameter of d = 10.0 (mm). Stresses are plotted onto the post-angioplasty con-
figuration at 13.3 (kPa). The luminal shape before angioplasty at 13.3 (kPa) is indicated by a
dashed curve.
may provide a 'virtual workbench' for the design and evaluation of interventional instruments,
such as stents. Moreover, the proposed model approach bears the potential for the determination
of lesion-type specific procedures, i.e. the definition of optimal procedural protocols for certain
plaque geometries and compositions with respect to certain target quantities such as the maxi-
mum luminal gain, intimal injury, medial overstretch etc. Another potential application evolves
from the fact that for the first time it is possible to investigate the change in the mechanical
environment caused by an interventional procedure on a layer-specific basis. This may allow a
better understanding of the corresponding biological response, and hence could be a potential
contributor for restenosis research.
Morphological data for such investigations may come from high-resolution magnetic res-
onance imaging of the arterial wall and plaque, which holds great promise for characterizing
atherosclerotic plaques and is, therefore, of huge interest for the medical community. MRI has
shown to discriminate between various tissue types involved in atherosclerosis in vitro (Toussaint
et al. (1995), Martinet a!. (1995), Shinnar eta!. (1999) and Rogers eta!. (2000)). Moreover, the
challenging task of in vivo wall imaging yielded promising results for carotid arteries (Yuan et
a!. (1995), Toussaint eta!. (1996) Luk-Pat eta!. (1999)), for the aorta (Fayad eta!. (2000b)),
and even for coronary arteries (Fayad et a!. (2000a)). Thus, it can be expected that in the near
future MR wall and plaque imaging will become clinical reality so that morphological input for
models as the proposed one are available. Clearly, the combination of mechanical analysis and
morphological data has an enormous potential to significantly improve current diagnostics and
interventional procedures.
Acknowledgements. I am indebted to many colleagues for their gentle encouragement and perspective
of arterial wall mechanics. I am pleased to acknowledge R. W. Ogden, Professor of Mathematics at the Uni-
versity of Glasgow, for reading and improving the manuscript, and for helpful discussions. Additionally, I
would like to acknowledge the essential support and the inspiring and detailed comments of my collabora-
tors. In particular, I wish to thank those of my collaborators who are involved in biomechanical research of
soft tissues, i.e. Ch.A.J. Schulze-Rauer, MD, T.Ch. Gasser, PhD, and my PhD students M. Stadler and M.
Auer. Financial support was provided by the Austrian Science Foundation under START-Award Y74-TEC.
This support is gratefully acknowledged.
References
Abe, H., Hayashi, K., and Sato, M., eds. (1996). Data Book on Mechanical Properties of Living Cells,
Tissues, and Organs. New York: Springer-Verlag.
American Heart Association. (2000). 2001 Heart and Stroke Statistical Update. Dallas, Texas: American
Heart Association.
Bathe, K.-J. (1996). Finite Element Procedures. Englewood Cliffs, New Jersey: Prentice-Hall.
Bazaraa, M. S., Sherali, H. D., and Shetty, C. M. (1993). Nonlinear Programming, Theory and Algorithms.
Chichester: John Wiley & Sons, 2nd edition.
Belytschko, T., Liu, W. K., and Moran, B. (2000). Nonlinear Finite Elements for Continua and Structures.
Chichester: John Wiley & Sons.
Berge!, D. H. (196la). The dynamic elastic properties of the arterial wall. J. Physiol. 156:458--469.
Berge!, D. H. (196lb). The static elastic properties of the arterial wall. J. Physiol. 156:445--457.
Bonet, J., and Wood, R. D. (1997). Nonlinear Continuum Mechanics for Finite Element Analysis. Cam-
bridge: Cambridge University Press.
Brezzi, F., and Fortin, M. (1991). Mixed and Hybrid Finite Element Methods. New York: Springer-Verlag.
Carmines, D. V., McElhaney, J. H., and Stack, R. (1991). A piece-wise non-linear elastic stress expression
of human and pig coronary arteries tested in vitro. J. Biomech. 24:899-906.
Castaneda-Zuniga, W. R. (1985). Pathophysiology of transluminal angioplasty. In Meyer, J., Erber!, R.,
and Rupprecht, H. J., eds., improvement of Myocardial Perfusion. Boston: Martinus Nijhoff Publisher.
138-141.
Chen, Y. L., and Fung, Y. C. (1973). Stress-history relations of rabbit mesentery in simple elongation. In
ASME 1973 Biomechanics Symposium, AMD-Vol. 2, 9-10. New York: American Society of Mechanical
Engineers.
Chung, J., and Hulbert, G. M. (1993). A time integration algorithm for structural dynamics with improved
numerical dissipation: The generalized a-method. J. Appl. Mech. 60:371-375.
Chuong, C. J., and Fung, Y. C. (1983). Three-dimensional stress distribution in arteries. ASME J. Biomech.
Eng. 105:268-274.
Coleman, B. D., and Gurtin, M. E. (1967). Thermodynamics with internal state variables. J. Chem. Phys.
47:597-613.
Coleman, B. D., and Noll, W. (1963). The thermodynamics of elastic materials with heat conduction and
viscosity. Arch. Rational Mech. Anal. 13:167-178.
Cox, R. H. (1976). Mechanics of canine iliac artery smooth muscle in vitro. Am. J. Physiol. 230:462--470.
Crisfield, M.A. (1991). Non-linear Finite Element Analysis ofSolids and Structures, Essentials, Volume 1.
Chichester: John Wiley & Sons.
180 G.A. Holzapfel
Crisfield, M.A. (1997). Non-linear Finite Element Analysis of Solids and Structures, Advanced Topics,
Volume 2. Chichester: John Wiley & Sons.
Deng, S. X., Tomioka, J., Debes, J. C., and Fung, Y. C. (1994). New experiments on shear modulus of
elasticity of arteries. Am. J Physiol. 266:Hl-H10.
Dennis, J. E., and Schnabel, R. B. (1983). Numerical Methods for Unconstrained Optimization and Non-
linear Equations. Englewood Cliffs, New Jersey: Prentice-Hall.
Eberlein, R., Holzapfel, G. A., and Schulze-Bauer, C. A. J. (2001). An anisotropic model for annulus tissue
and enhanced finite element analyses of intact lumbar disc bodies. Comput. Meth. Biomech. Biomed.
Engr. 4:209-230.
Fayad, Z. A., Fuster, V., Fallon, J. T., Jayasundera, T., Worthley, S. G., Helft, G., Aguinaldo, J. G., Badimon,
J. J., and Sharma, S. K. (2000a). Noninvasive in vivo human coronary artery lumen and wall imaging
using black-blood magnetic resonance imaging. Circ. 102:506--510.
Fayad, Z. A., Nahar, T., Fallon, J. T., Goldman, M., Aguinaldo, J. G., Badimon, J. J., Shinnar, M., Chesebro,
J. H., and Fuster, V. (2000b ). In vivo magnetic resonance evaluation of atherosclerotic plaques in the
human thoracic aorta: A comparison with transesophageal echocardiography. Circ. 101:2503-2509.
Fischman, D. L., Leon, M. B., Bairn, D. S., Schatz, R. A., Savage, M. P., Penn, 1., Detre, K., Veltri, L.,
Ricci, D., Nobuyoshi, M., Cleman, M. W., Heuser, R. R., Almond, D., Teirstein, P. S., Fish, R. D.,
Colombo, A., Brinker, J., Moses, J., Shaknovich, A., Hirshfeld, J., Bailey, S., Ellis, S., Rake, R., and
Goldberg, S. (1994). A randomized comparison of coronary-stent placement and balloon angioplasty
in the treatment of coronary artery disease. Stent restenosis study investigators. New Engl. J Med.
331:496--501.
Flory, P. J. ( 1961 ). Thermodynamic relations for highly elastic materials. Trans. Faraday Soc. 57:829-838.
Fung, Y C., and Sobin, S. S. (1981 ). The retained elasticity of elastin under fixation agents. J Biomech.
Engr. 103:121-122.
Fung, Y C., Fronek, K., and Patitucci, P. (1979). Pseudoelasticity of arteries and the choice of its mathe-
matical expression. Am. J Physiol. 237:H620--H631.
Fung, Y C. ( 1971 ). Stress-strain-history relations of soft tissues in simple elongation. In Fung, Y C., Per-
rone, N., and Anliker, M., eds., Biomechanics: Its Foundations and Objectives. New Jersey: Prentice-
Hall, Inc., Englewood Cliffs. 181-208. Chapter 7.
Fung, Y C. (1993). Biomechanics. Mechanical Properties of Living Tissues. New York: Springer-Verlag,
2nd edition.
Gasser, T. C., and Holzapfel, G. A. (2001). A rate-independent elastoplastic constitutive model for fiber-
reinforced composites at finite strains: Continuum basis, algorithmic formulation and finite element
implementation. Comput. Mech. in press.
Gasser, T. C., Schulze-Bauer, C. A. J., and Holzapfel, G. A. (2002). A three-dimensional finite element
model for arterial clamping. ASME J Biomech. Eng. 124:355-363.
Goudreau, G. L., and Taylor, R. L. (1972). Evaluation of numerical integration methods in elastodynamics.
Comput. Meth. Appl. Mech. Engr. 2:69-97.
Gow, B. S., and Hadfield, C. D. (1979). The elasticity of canine and human coronary arteries with reference
to postmortem changes. Circ. Res. 45:588-594.
Harvey, J. G., and Gough, M. H. (1981). A comparison of the traumatic effects of vascular clamps. Br. J
Surg. 68:267-272.
Hayashi, K., Kamiya, A., and Ono, K., eds. (1996). Biomechanics- Functional Adaptation and Remodel-
ing. Tokyo: Springer-Verlag.
Hayashi, K. (1993). Experimental approaches on measuring the mechanical properties and constitutive
laws of arterial walls. J Biomech. Engr. 115:481--488.
Herrmann, L. R., and Peterson, F. E. (1968). A numerical procedure for viscoelastic stress analysis. in
Proceedings 7th Meeting ofICRPG Mechanical Behavior Working Group, Orlando.
Hibbitt, Karlsson, and Sorensen., eds. (1998). ABAQUS/Standard Users's Manual, Version 5.8. Hibbitt,
Karlsson and Sorensen, Inc.
Hilber, H. M., Hughes, T. J. R., and Taylor, R. L. (1977). Improved numerical dissipation for time integra-
tion algorithms in structural dynamics. Earthquake Eng. Struc. 5:282-292.
Hinton, E., Rock, T., and Zienkiewicz, 0. C. (1976). A note on mass lumping and related processes in the
finite element method. Earthquake Eng. Struc. 4:245-249.
Holzapfel, G. A., and Gasser, T. C. (2001). A viscoelastic model for fiber-reinforced composites at finite
strains: Continuum basis, computational aspects and applications. Comput. Meth. Appl. Mech. Engr.
190:43 79--4403.
Holzapfel, G. A., and Weizsiicker, H. W. (1998). Biomechanical behavior of the arterial wall and its
numerical characterization. Camp. Bioi. Med. 28:377-392.
Holzapfel, G. A., Gasser, T. C., and Ogden, R. W. (2000a). A new constitutive framework for arterial wall
mechanics and a comparative study of material models. J. Elasticity 61: 1--48.
Holzapfel, G. A., Schulze-Sauer, C. A. J., and Stadler, M. (2000b). Mechanics of angioplasty: Wall,
balloon and stent. In Casey, J., and Bao, G., eds., Mechanics in Biology. New York: The American
Society of Mechanical Engineers (ASME). AMD-Vol. 242/BED-Vol. 46, pp. 141-156.
Holzapfel, G. A., Gasser, T. C., and Stadler, M. (2002a). A structural model for the viscoelastic behavior of
arterial walls: Continuum formulation and finite element analysis. Eur. J. Mech. A/Solids 21 :441--463.
Holzapfel, G. A., Stadler, M., and Schulze-Bauer, C. A. J. (2002b ). A layer-specific 3D model for the finite
element simulation of balloon angioplasty using MR imaging and mechanical testing. Ann. Biomed.
Engr. 30:753-767.
Holzapfel, G. A. (1996). On large strain viscoelasticity: Continuum formulation and finite element appli-
cations to elastomeric structures. Int. J. Numer. Meth. Engr. 39:3903-3926.
Holzapfel, G. A. (2000). Nonlinear Solid Mechanics. A Continuum Approach for Engineering. Chichester:
John Wiley & Sons.
Holzapfel, G. A. (2001). Biomechanics of soft tissue. In Lemaitre, J., ed., The Handbook of Materials
Behavior Models. Volume IlL Multiphysics Behaviors, Chapter 10, Composite Media, Biomaterials,
1049-1063. Boston: Academic Press.
Hoschek, J., and Lasser, D., eds. (1993). Fundamentals of Computer Aided Geometric Design. Wellesly:
A. K. Peters Ltd.
Hughes, T. J. R., and Pister, K. S. (1978). Consistent linearization in mechanics of solids and structures.
Comput. & Structures 8:391-397.
Hughes, T. J. R., Liu, W. K., and Brooks, A. (1979). Finite element analysis of incompressible viscous
flows by the penalty function formulation. J. Comput. Phys. 30: 1-60.
Hughes, T. J. R. (2000). The Finite Element Method: Linear Static and Dynamic Finite Element Analysis.
New York: Dover.
Humphrey, J. D. (1995). Mechanics of the arterial wall: Review and directions. Critical Reviews in
Biomed. Engr. 23:1-162.
Humphrey, J. D. (2002). Cardiovascular Solid Mechanics. Cells, Tissues, and Organs. New York:
Springer-Verlag.
Jackiewicz, T. A., McGeachie, J. K., and Tennant, M. (1996). Structural recovery of small arteries follow-
ing clamp injury: a light and electron microscopic investigation in the rat. Microsurgery 17:674-680.
Langewouters, G. J., Wesseling, K. H., and Goedhard, W. J. A. (1984). The static elastic properties of
45 human thoracic and 20 abdominal aortas in vitro and the parameters of a new model. J. Biomech.
17:425--435.
Learoyd, B. M., and Taylor, M. G. (1966). Alterations with age in the viscoelastic properties of human
arterial walls. Circ. Res. 18:278-292.
Lubliner, J. (1990). Plasticity Theory. New York: Macmillan Publishing Company.
Luenberger, D. G. (1984). Linear and Nonlinear Programming. Reading, Massachusetts: Addison-Wesley
Publishing Company.
Luk-Pat, G. T., Gold, G. E., Olcott, E. W., Hu, B. S., and Nishimura, D. G. (1999). High-resolution
three-dimensional in vivo imaging of atherosclerotic plaque. Magn. Reson. Med. 42:762-771.
182 G.A. Holzapfel
Macaya, C., Serruys, P. W., Ruygrok, P., Suryapranata, H., Mast, G., Klugmann, S., Urban, P., den Heijer,
P., Koch, K., Simon, R., Morice, M. C., Crean, P., Bonnier, H., Wijns, W., Danchin, N., Bourdonnec, C.,
and Morel, M. A. ( 1996). Continued benefit of coronary stenting versus balloon angioplasty: One-year
clinical follow-up ofBenestent trial. Benestent Study Group. J. Am. Call. Cardia!. 27:255-61.
Maltzahn, W.-W. V., and Warriyar, R. G. (1984). Experimental measurments of elastic properties of media
and adventitia of bovine carotid arteries. J. Biomech. 17:839-847.
Margovsky, A. 1., Lord, R. S. A., Meek, A. C., and Bobryshev, Y. V. (1997). Artery wall damage and
platelet uptake from so-called atraumatic arterial clamps: An experimental study. Cardia!. Surg. 5:42-
47.
Mario, C. D., Gil, R., Camenzind, E., Ozaki, Y., von Birgelen, C., Umans, V., de Jaegere, P., de Feyter, P. J.,
Roelandt, J. R. T. C., and Serruys, P. W. (1995). Quantitative assessment with intracoronary ultrasound
of the mechanism of restenosis after percutaneous transluminal coronary angioplasty and directional
coronary atherectomy. Am. J. Cardia!. 75:772-777.
Martin, A. J., Gotlieb, A. 1., and RM, R. M. H. (1995). High-resolution MR imaging of human arteries. J.
Magn. Reson. Imaging 5:93-100.
Matthies, H., and Strang, G. (1979). The solution of nonlinear finite element equations. Int. J. Numer.
Meth. Engr. 14:1613-1626.
Nagtegaal, J. C., Parks, D. M., and Rice, J. R. (1974). On numerically accurate finite element solutions in
the fully plastic range. Comput. Meth. Appl. Mech. Engr. 4:153-177.
Newmark, N. M. (1959). A method of computation for structural dynamics. J. Eng. Mech. Division, Proc.
Am. Soc. of Civil Eng. 85:67-94.
Nimni, M. E., and Harkness, R. D. (1988). Molecular structure and functions of collagen. In Nimni, M. E.,
ed., Collagen. Boca Raton, FL: CRC Press. 3-35.
Nimni, M. E., ed. (1988). Collagen. 4 Vols: 1. Biochemistry; 2. Biochemistry and Biomechanics; 3.
Biotechnology; 4. Molecular Biology. Boca Raton, FL: CRC Press.
Oden, J. T., and Carey, G. F. (1984). Finite Elements: Mathematical Aspects, Volume IV. Englewood
Cliffs, N.J.: Prentice-Hall.
Ogden, R. W., and Schulze-Bauer, C. A. J. (2000). Phenomenological and structural aspects of the me-
chanical response of arteries. In Casey, J., and Bao, G., eds., Mechanics in Biology. New York: The
American Society of Mechanical Engineers (ASME). AMD-Vol. 242/BED-Vol. 46, pp. 125-140.
Ogden, R. W. (1978). Nearly isochoric elastic deformations: Application to rubberlike solids. J. Mech.
Phys. Solids 26:37-57.
Ogden, R. W. (1997). Non-linear Elastic Deformations. New York: Dover.
Patel, D. J., and Fry, D. L. (1969). The elastic symmetry of arterial segments in dogs. Circ. Res. 24:1-8.
Piegel, L.A., and Tiller, W. (1997). The NURBS Book. New York: Springer-Verlag, 2nd edition.
Rachev, A., Manoach, E., Berry, J., and Moore Jr., J. E. (2000). A model of stress-induced geometri-
cal remodeling of vessel segments adjacent to stents and artery/graft anastomoses. J. Theoret. Bioi.
206:429--443.
Rhodin, J. A. G. (1980). Architecture of the vessel wall. In Bohr, D. F., Somlyo, A. D., and Sparks, H. V.,
eds., Handbook of Physiology, The Cardiovascular System, Volume 2. Bethesda, Maryland: American
Physiologial Society. 1-31.
Richardson, P. D., Davies, M. J., and Born, G. V. R. (1989). Influence of plaque configuration and stress
distribution on fissuring of coronary atherosclerotic plaques. Lancet 2(8669):941-944.
Riks, E. (1984). Some computational aspects of stability analysis of nonlinear structures. Comput. Meth.
Appl. Mech. Engr. 47:219-260.
Roach, M. R., and Burton, A. C. (1957). The reason for the shape of the distensibility curve of arteries.
Canad. J. Biochem. Physiol. 35:681-690.
Rodriguez, E. K., Hoger, A., and McCulloch, A. D. (1994). Stress-dependent finite growth in soft elastic
tissues. J Biomech. 27:455-467.
Rogers, W. J., Prichard, J. W., Hu, Y. L., Olson, P. R., Benckart, D. H., Kramer, C. M., Vida, D. A., and
Reichek, N. (2000). Characterization of signal properties in atherosclerotic plaque components by
intravascular MRI. Arterioscl. Thromb. and Vase. Bioi. 20:1824-1830.
Roy, C. S. (1880-82). The elastic properties of the arterial wall. J Physiol. 3:125-159.
Schultze-Jena, B. S. (1939). Uber die schraubenformige Struktur der Arterienwand. Gegenbauers Mor-
pho/. Jahrbuch 83:230--246.
Schulze-Bauer, C. A. J., Morth, C., and Holzapfel, G. A. (2001). Passive biaxial mechanical response of
aged human iliac arteries. ASME J. Biomech. Eng. in press.
Schulze-Bauer, C. A. J., Regitnig, P., and Holzapfel, G. (2002). Mechanics of the human femoral adventitia
including high-pressure response. Am. J. Physiol.. Heart Circ. Physiol. 282:H2427-2440.
Schweizerhof, K. H., and Ramm, E. (1984). Displacement dependent pressure loads in nonlinear finite
element analysis. Comput. & Structures 18:1099-1114.
Shinnar, M., Fallon, J. T., Wehrli, S., Levin, M., Dalmacy, D., Fayad, Z. A., Badimon, J. J., Harrington, M.,
Harrington, E., and Fuster, V. (1999). The diagnostic accuracy of ex vivo mri for human atherosclerotic
plaque characterization. Arterioscl. Thromb. and Vase. Bioi. 19:2756--2761.
Silver, F. H., Christiansen, D. L., and Buntin, C. M. (1989). Mechanical properties of the aorta: A review.
Critical Reviews in Biomed. Engr. 17:323-358.
Sima, J. C., and Hughes, T. J. R. (1998). Computational Inelasticity. New York: Springer-Verlag.
Sima, J. C., and Taylor, R. L. (1991). Quasi-incompressible finite elasticity in principal stretches.
Continuum basis and numerical algorithms. Comput. Meth. Appl. Mech. Engr. 85:273-310.
Sima, J. C., Taylor, R. L., and Pister, K. S. (1985). Variational and projection methods for the volume
constraint in finite deformation elasto-plasticity. Comput. Meth. Appl. Mech. Engr. 51:177-208.
Skalak, R., Zargaryan, S., Jain, R. K., Netti, P. A., and Hoger, A. (1996). Compatibility and the genesis of
residual stress by volumetric growth. J. Math. Bioi. 34:889-914.
Slayback, J. B., Bowen, W. W., and Hinshaw, D. B. (1976). Intimal injury from arterial clamps. Am. J.
Surg. 132:183-188.
Spencer, A. J. M. (1984). Constitutive theory for strongly anisotropic solids. In Spencer, A. J. M., ed.,
Continuum Theory of the Mechanics of Fibre-Reinforced Composites. Wien: Springer-Verlag. 1-32.
CISM Courses and Lectures No. 282, International Centre for Mechanical Sciences.
Stary, H. C., Blankenhorn, D. H., Chandler, A. B., Glagov, S., Jr. Insull, W., Richardson, M., Rosenfeld,
M. E., Schaffer, S. A., Schwartz, C. J., Wagner, W. D., and Wissler, R. W. (1992). A definition of
the intima of human arteries and of its atherosclerosis-prone regions. A report from the committee on
vascular lesions of the council on arteriosclerosis, American Heart Association. Circ. 85:391-405.
Staubesand, J. (1959). Anatomie der BlutgefaBe. I. Funktionelle Morphologie der Arterien, Venen und
arterio-venosen Anastomosen. In Ratschow, M., ed., Angiology. Stuttgart: Thieme. Chapter 2, 23-82.
Taber, L. (1995). Biomechanics of growth, remodelling, and morphognesis. Appl. Mech. Rev. 48:487-543.
Taber, L. A. (1998). A model for aortic growth based on fluid shear and fiber stress. J. Biomech. Engr.
120:348-354.
Tanaka, T. T., and Fung, Y. C. (1974). Elastic and inelastic properties of the canine aorta and their variation
along the aortic tree. J. Biomech. 7:357-370.
Taylor, R. L., Pister, K. S., and Goudreau, G. L. (1970). Thermomechanical analysis of viscoelastic solids.
Int. J. Numer. Meth. Engr. 2:45-59.
Taylor, R. L. (2000). FEAP- A Finite Element Analysis Program - Version 7. 3. University of California at
Berkeley.
The, S. H. K., Gussenhoven, E. J., Zhong, Y., Li, W., van Egmond, F., and Pieterman, H. (1992). Effect of
balloon angioplasty on femoral artery evaluated with intravascular ultrasound imaging. Circ. 86:483-
493.
184 G.A. Holzapfel
Toussaint, J. F., Southern, J. F., Fuster, V., and Kantor, H. L. (1995). T2 contrast for nrnr characterization
of human atherosclerosis. Arterioscl. Throm b. and Vase. Bioi. 15: 1533-1542.
Toussaint, J. F., LaMuraglia, G. M., Southern, J. F., Fuster, V., and Kantor, H. L. (1996). Magnetic reso-
nance images lipid, fibrous, calcified, hemorrhagic, and thrombotic components of human atheroscle-
rosis in vivo. Circ. 94:932-938.
Vaishnav, R.N., and Vossoughi, J. (1983). Estimation of residual strains in aortic segments. In Hall, C. W.,
ed., Biomedical Engineering II: Recent Developments. New York: Pergamon Press. 330-333.
Valanis, K. C. (1972). Irreversible Thermodynamics ofContinuous Media, Internal Variable Theory. Wien:
Springer-Verlag.
Washizu, K. (1982). Variational Methods in Elasticity and Plasticity. Oxford: Pergamon Press. 3rd
edition.
Weizsiicker, H. W., and Pinto, J. G. (1988). Isotropy and anisotropy of the arterial wall. J Biomech.
21:477-487.
Wolinsky, H., and Glagov, S. (1967). A lamellar unit of aortic medial structure and function in mammals.
Circ. Res. 20:90-111.
Woo, S. L. Y., Simon, B. R., Kuei, S.C., and Akeson, W. H. (1979). Quasi-linear viscoelastic properties of
normal articular cartilage. J Biomech. Engr. 102:85-90.
Wood, W. L., Bossak, M., and Zienkiewicz, 0. (1981). An alpha modification ofNewmark's method. Int.
J Numer. Meth. Engr. 15:1562-1566.
Wood, W. L. (1990). Practical Time-stepping Schemes. Oxford: Clarendon Press.
Wriggers, P. (2001). Nichtlineare Finite-Element-Methoden. Berlin: Springer-Verlag.
Xie, J., Zhou, J., and Fung, Y. C. ( 1995). Bending of blood vessel wall: Stress-strain laws of the intima-
media and adventitia layers. J Biomech. Engr. 117:136-145.
Yu, Q., Zhou, J., and Fung, Y. C. (1993). Neutral axis location in bending and Young's modulus of different
layers of arterial wall. Am. J Physiol. 265 :H52-H60.
Yuan, C., Murakami, J. W., Hayes, C. E., Tsuruda, J. S., Hatsukami, T. S., Wildy, K. S., Ferguson, M.S., and
Strandness, D. E. (1995). Phased-array magnetic resonance imaging of the carotid artery bifurcation:
Preliminary results in healthy volunteers and a patient with atherosclerotic disease. J Magn. Reson.
Imaging 5:561-565.
Zienkiewicz, 0. C., and Taylor, R. L. (2000). The Finite Element Method. Solid Mechanics, Volume 2.
Oxford: Butterworth Heinemann, 5th edition.
Intracranial Saccular Aneurysms
J.D. Humphrey
Biomedical Engineering, Texas A&M University, College Station, TX77843-3120, USA

E-mail: jhumphrey@tamu.edu, Home Page: http://biomed.tamu.edu/faculty/humphrey/
Abstract. It has long been accepted that mechanical factors play a critical role in the natu-
ral history of intracranial saccular aneurysms - their pathogenesis, enlargement, and rup-
ture. Nevertheless, until very recently, biomechanical analysis has been surprisingly scant.
In this chapter, we see how nonlinear elasticity, membrane theory, nonlinear dynamics, and
nonlinear finite element analyses can be used to increase our understanding of the mechan-
ics of saccular aneurysms. Much has been learned, but much remains to be accomplished,
particularly in the area of the mechanics of biological growth and remodeling. Hence, it
is also hoped that this chapter will encourage new investigators to study the mechanics of
aneurysms.
1 Introduction
Intracranial saccular aneurysms are focal dilatations of the arterial wall that occur in or near bi-
furcations within the circle of Willis, the primary network of vessels that supply blood to the
brain. Two-to-five percent of the general population in the Western World likely harbors a sac-
cular aneurysm. Histopathological and clinical studies reveal further that these lesions are more
common in women than in men, that they are more prevalent in the 4th to 6th decades of life,
and that they occur predominantly in the anterior and middle portions of the cerebral vascula-
ture. Moreover, these lesions often remain asymptomatic until rupture, whichis the leading cause
of spontaneous subarachnoid hemorrhage (the mean age at rupture being around 52 years old).
Despite recent advances in neurosurgery and neuroradiology, 35-50% of subarachnoid hemor-
rhage patients die whereas many of the survivors suffer severe functional deficits. There is a
need, therefore, to improve our understanding and treatment of this devastating disease. For the
purposes herein, it is important to note that it is widely thought that rupture occurs when the
hemodynamically-induced wall stress exceeds wall strength. Yet, there have been few rigorous
biomechanical analyses and it is hoped that this brief presentation (Sections 1-4) will encourage
this much needed research. For more details, see Humphrey (2002), on which this chapter was
based.
1.1 Natural History

The natural history of saccular aneurysms can be thought to consist of three phases: pathogenesis,
enlargement, and rupture. There are many different theories with regard to the initiation of sac-
cular aneurysms but little general agreement (see, e.g., Sekhar and Heros, 1981). Nevertheless,
it appears that unique structural features of the cerebral vasculature, risk factors, genetics, and
hemodynamics each contribute to the pathogenesis. In particular, cerebral arteries do not have an
external elastic lamina, they have sparse medial elastin, they lack supporting perivascular tissue,
186 J.D. Humphrey
and they appear to have structural irregularities at the apex of their bifurcations. These features
may render the cerebral artery susceptible to a local weakening of the wall under the persistent
action of hemodynamic loads. It is thought, for example, that the internal elastic lamina and mus-
cular media must become markedly fragmented or disappear in order for a saccular aneurysm to
form. In addition to hypertension, risk factors may include heavy alcohol consumption, cigarette
smoking, and the long term use of analgesics, oral contraceptives, or cocaine (see, de la Monte
et al., 1985, Weibers et al., 1998). Increased familial incidence in some populations suggests that
genetics likewise play an important role. There is a pressing need for much more research on the
role of genetics and hemodynamics in the pathogenesis, however.
Beginning as a small outpouching, or dilatation, of the arterial wall, saccular aneurysms may
expand to over 30 mm in diameter and become complex in shape and constitution. Unfortu-
nately, little is known about the mechanisms by which this enlargement occurs, or its time-course.
Among other hypotheses, it has been suggested that lesions expand rapidly due to structural in-
stabilities, as, for example, a limit point instability or resonance. These two hypotheses are ex-
amined in Section 3 based on more recent nonlinear analyses and shown to be unlikely, at least
for particular classes of lesions. It seems, therefore, that aneurysms enlarge via a growth and
remodeling response, although again there has been little careful analysis along this line.
The final, and often fatal, phase of the natural history of a saccular aneurysm is its rupture. In
this context, rupture may imply one of two outcomes: a catastrophic tearing of a portion of the
lesion and significant bleeding, or a small bleed that is sealed by a fibrin patch and followed by
the formation of an intraluminal or intramural thrombus. In the latter, the lesion is likely more
susceptible to subsequent enlargement or catastrophic rupture. Regardless, the mean annual risk
of rupture is between~ 0.1 to 1%; thus, the prognosis 10 or more years out can be good (Wiebers
et al., 1998). Rupture usually occurs at the fundus (or pole) even though the neck may be thinner.
Moreover, in the case of coexisting aneurysms, the larger one usually ruptures first, or if of nearly
the same size, the proximal one usually ruptures first.
1.2 Histopathology
Among others, Stehbens (1990) provides a nice review of lesion histology. Briefly, it appears
that 'non-complicated' saccular aneurysms are often transparent, consisting primarily of an en-
dothelium and a thin remnant of the adventitia. Hence, the main structural constituent is type I
collagen, orientations of which are discussed below. There is some evidence that the endothelial
cells are slightly thickened, with an increased endoplasmic reticulum suggesting a heightened
synthetic activity- probably the deposition of type IV collagen as well as adhesion molecules
such as fibronectin and laminin. This would be consistent, of course, with an increased replica-
tion of endothelial cells to cover the expanding inner surface of a lesion. Sparse, irregular smooth
muscle cells may play a role in regulating the intramural collagen, but this is likely accomplished
primarily by fibroblasts. Of course, the regulation of collagen results from a competition between
deposition and degradation. A number of recent papers (e.g., Chyatte and Lewis, 1997, Bruno et
al. 1998) implicate an increased protease activity within aneurysms. That is, it appears that the
activity of matrix metalloproteinases (MMPs) and tissue inhibitors of such (TIMPs) is signifi-
cantly higher in aneurysms than in nearby cerebral arteries; this may be the reason for a decrease
in types III and V collagen in aneurysms' (Hegedus, 1984, Gaetani eta!., 1997). Likewise, an
1 The latter is also likely due in part to the absence of smooth muscle cells in the aneurysms
Intracranial Saccular Aneurysms 187
increased turnover of type I collagen may result in more immature collagen, characterized by
less stable cross-links and thus attenuated strength (Canham et al., 1999).
As a lesion continues to enlarge, it appears that intramural macrophages may accumulate
some lipids, particularly on lateral portions of the sac. The associated intimal thickening is often
the first sign of the development of atherosclerosis, advanced forms of which are characterized
by proliferating smooth muscle cells, lymphocytes, and activated macrophages (Kosierkiewicz et
al., 1994). Because advanced atherosclerosis is typically found only in the larger aneurysms, this
suggests that atherosclerosis is often a response to a prior injury rather than part of the pathogene-
sis. Nonetheless, atherosclerosis may contribute to the continued expansion and ultimate rupture
oflesions due to its inflammatory nature. Finally, larger aneurysms are often complicated further
by prior leaks and associated repairs, which are evidenced by the presence of thrombi.
In addition to the type and volume fraction of the collagen, its cross-link density and orien-
tations within the wall of the lesion contribute significantly to the overall mechanical properties.
P.B. Canham and colleagues at the University of Western Ontario in Canada have reported the
best data on these two critical parameters in human saccular aneurysms. Briefly, they use two
microscopic techniques that exploit the birefringent properties of collagen: the Senarmont com-
pensator allows them to measure the intensity of the birefringence, which is a function of the
cross-link density and size of the collagen fibers, whereas a universal stage attachment on a
polarizing microscope allows them to measure 3-D orientations of collagen within the wall. Ori-
entation is prescribed by two angles, azimuth (in-plane direction) and elevation (out-of-plane),
azimuth being most important in tangential sections. Canham et al. (1991) have shown that they
can resolve fiber orientations to within 0. 7° in 3-5 J.Lm serial sections that are stained with 0.05
to 1% picrosirius red, which enhances the birefringence.
Among other findings, Whittaker et al. (1988) suggested that local decreases in wall strength
near rupture sites correspond to the presence of immature type I collagen rather than to previ-
ously speculated reasons, such as the degradation of mature type I collagen, a prevalence of type
III collagen, or a decreased proteoglycan concentration (note: they did not assay protease activity
or the presence of macrophages ). Moreover, they reported regional variations in collagen orien-
tation, although no trends were found, and that the collagen in the outer portion of the wall was
more mature but less ordered than that in the inner wall. The latter suggests that the deposition
of new collagen occurs in the inner wall. Canham et al. (1996) reported further that aneurysmal
collagen has little waviness when perfusion-fixed at physiologic pressures, that it follows great
circle trajectories (i.e., trace a circular path in a plane that passes through the center of the lesion),
and that it is organized in 7 or 8 distinctive layers, each of which consists of nearly parallel fibers
(±8°). In general, the layer-to-layer orientation changed abruptly, not continuously, often to an
opposite direction. In the region of the fundus on a small spherical lesion, they report that taken
together, collagen from all the layers spanned the full range of azimuthal angles - this implies an
isotropic response near the fundus as required by axisymmetry, and hence a consistency between
the often postulated mechanics and observed microstructure. Indeed, Canham et al. (1999) stated
that ' .. .in general aneurysms are well designed and mechanically stable.'
Finally, MacDonald et al. (2000) reported further that thicker-walled regions appeared to be
dominated by weaker collagen in contrast to the prevalence of stronger collagen in the thinner-
walled regions. This may be due in part to the greater weakness of the newly deposited collagen,
which again, tends to be in the inner portion of the wall. These observations were based on the
intensity ofthe birefringence across the wall in one lesion; the intensity, and hence wall strength,
188 J.D. Humphrey
increased from the inner to the outer wall although the strength at the outer wall did not reach
that of the adventitial layer in the nearby parent vessel. Overall, therefore, they suggest that ' ... the
wall of an aneurysm may include a mix of newly polymerized collagen fibers and reorganized
fibers from the earlier, developing lesion.'
Because the microstructure of the wall (and thus properties) is but one determinant of wall
stress (another being 3-D geometry, including wall thickness), quantitative histology alone can-
not definitively identify determinants of rupture-potential. There is a clear need to combine ge-
ometrical, mechanical, and sophisticated histological methods, such as those developed by Can-
ham and colleagues, to understand more completely the mechanics of the aneurysmal wall.
1.3 Mechanical Properties
Prior Reports. Scott et al. (1972) performed in vitro pressure-volume tests on seven human sac-
cular aneurysms and 16 normal intracranial arteries. The aneurysms were assumed to be perfect
spheres, having defortned volumes of 4na 3 /3, from which the deformed radii a were calcu-
lated. They showed that aneurysms exhibit a nonlinear behavior over finite strains, and reported
that they were stiffer than normal vessels; moreover, two aneurysms exhibited critical breaking
stresses on the order of u c = 2 to 3 MPa. These findings were based on graphical comparisons
of tension (T = Pa/2 in spherical aneurysms and T = Pain cylindrical arteries, where Pis
the distending pressure and a the deformed inner radius) versus 'strain' data. Note, however, that
the measures of strain were different for the two types of specimens: A2 - 1 for the aneurysms,
which is twice the circumferential Green strain (and calculated based on changes in surface area,
(4na 2 - 4n A2 )/4n A2 , where A is the unloaded radius), and A- 1 for the arteries, which is the
linearized strain (calculated based on changes in circumference, (2na- 2nA)/2nA). In each
case, the stretch ratio A = a/ A, the ratio of radii in the pressurized and unloaded configurations.
Yet, the unloaded configuration in the artery is not stress-free whereas that in the membranous
aneurysm is stress-free by definition (i.e., a membrane cannot support a compressive stress and
thus cannot support a residual stress which must be self-equilibrating by definition). Clearly,
therefore, the qualitative conclusions offered by Scott et al. (1972) must be viewed carefully,
which reinforces the need for theory to guide the performance and interpretation of experiments
so that one knows what to measure and why. Finally, because they measured global volumes, not
local strains, the results are averaged and effectively one-dimensional, and thus not sufficient for
quantifying multiaxial constitutive relations.
Steiger et al. (1986) reported results from uniaxial extension tests on thin strips of tissue
excised from six human saccular lesions. An important finding was that lesion behavior differed
at the fundus and neck: tearing occurred at A = 1.37 and u c = 0.5 MPa in the fundus and
A= 1.57 and Uc = 1.21 MPa in the neck. That strain, not stress, was a more consistent measure
offailure is consistent with results on the failure of normal vessels (see Humphrey, 1995, pp. 77-
79). Nevertheless, Steiger's study is limited because the data were reduced using the linearized
measure of strain defined over the overall length of the specimen rather than a central gage length.
Moreover, the reported data did not discriminate between the meridional and circumferential
behaviors, and being 1-D, they are not complete for evaluating a 3-D constitutive relation.
Toth et al. (1998) recently reported similar uniaxial data from 22 human aneurysms, includ-
ing 17 harvested at surgery. Characteristics of the latter group include a mean diameter of 11.6
mm (ranging from 5 to 23 mm), a mean age of 47 years old (range from 32 to 63), that 12 of
the 17 surgical specimens were from females, and that 11 of the 17 patients had a subarachnoid
hemorrhage. Similarly to Steiger et al. ( 1986), they found that (circumferentially oriented) spec-
imens from the fundus tore at lower strains (i.e., A = 1.23) than those from the neck (.A = 1.55).
Moreover, they found that the strength near the fundus was greater in the meridional than in
the circumferential direction. Unfortunately, most results are presented in terms of moduli for a
Kelvin-Voigt linear viscoelastic model. The use of a linearized measure of strain is clearly inap-
propriate given their report of strains of up to 70%. Remarkably, these three studies represent the
entirety of the mechanical data on human aneurysms up to 1999.
Although Scott's data are not sufficient for detailed quantification ofmultiaxial behavior, in-
cluding anisotropy and heterogeneity, they appear to be the best available on human lesions. Kyr-
iacou and Humphrey (1996) showed that these data are well described by a Fung-type pseudo-
strain-energy function w, which is defined per undeformed surface area consistent with the direct
membrane approach (see Section 2 or Humphrey, 1998). Thus, consider a w of the form
w = c(eQ- 1), Q = c1Ei1 + c2Ei2 + 2c3EuE22, (1)

where EAB are the principal (in-plane) Green strains and c and ci are material parameters. For
the inflation of a perfectly spherical membrane (as assumed by Scott et al., 1972), the 2-D defor-
mation gradient tensor F is given by F = diag [A, A], where A = a/ A as above, and the principal
Green strains are E 11 = E 22 = ( .A 2 - 1) /2. Given a general constitutive relation for a membrane
(Humphrey, 1998),
1 8w T 1 OW
T =- - F · -8
det F E
·F --+ Taf3 = -Fa"'Ff3Ll-
J2n ~
-
8EsLl (a,,B,S,Ll = 1,2), (2)
where summation on repeated indices is implied per the usual Einstein convention. With c 2 c1 =
=
due to the implicit assumption of in-plane isotropy in Scott's analysis, one obtains from equations
( 1) and (2) the following expression for the uniform tension T1 T2 = T:
T = cFeQ(.A 2 -1), Q = ~F(.A 2 - 1) 2, (3)
where r := (c1 + c3). Kyriacou and Humphrey (1996) determined the best-fit values of the
r
two independent material parameters c and via a Marquardt-Levenberg regression of the data
presented by Scott et al. (1972), that is, by minimizing the sum-of-the-square of the error e
between the calculated and measured uniform stress resultants,
e =I: m (
creQ(.A 2 - 1)- p2a )2 , (4)
k=l k
where P is the measured distension pressure, Pa/2 the Laplace solution for the inflation of a
spherical membrane and m the number of equilibrium configurations (i.e., pressurized states).
The best-fit values were c = 0.88 N/m and r = (c 1 + c3) = 12.99, which yielded a reasonable
fit to data. Despite this good fit, the inadequacy of the data is evident: they do not allow sepa-
rate determination of c1 and c2, which embody the material symmetry, they do not separate the
contributions due to c1 and c3, and they do not provide information on possible heterogeneities.
=
Based on results on arteries reported by Fung and colleagues, Kyriacou and Humphrey (1996)
assumed that c3 ~ ci/10, and hence used the following values: c = 0.88 N/m, c 1 c2 = 11.82
and c3 = 1.18 in subsequent simulations.
190 J.D. Humphrey
An Alternate Approach. There are five general steps in the formulation of a constitutive relation
(DEICE): Delineating general characteristics, Establishing a theoretical framework, Identifying
a specific form of the constitutive relation, Calculating best-fit values of the material parameters,
and Evaluating the predictive capability of the final relation (Humphrey, 2002). Albeit yet to be
employed, Hsu et al. (1994, 1995) presented both a new theoretical framework and a multiaxial
experimental system for accomplishing steps 2-5 for thin, axisymmetric, non-complicated sac-
cular aneurysms that consist primarily of uniformly distributed collagen and thus exhibit a pseu-
doelastic behavior. 2 Based on the observations of Scott et al. (1972) and Steiger et al. (1986), the
nonlinear theory of elastic membranes is an appropriate theoretical framework. In Section 2, it is
shown that the two equilibrium equations for the axisymmetric inflation of a membrane can be
written as
(5)
which admit closed-form solutions independent of the specific constitutive behavior. That is, the
principal stress resultants T 1 and T2 are given by
(6)
where P is again the distension pressure and "'a

are local principal curvatures in the deformed
configuration (a = 1, 2 denotes the meridional and circumferential directions, respectively). Not-
ing that the pressure and curvatures are both experimentally measurable (curvatures by measuring
a single profile of the lesion r = r(s), where sis the arc length in the deformed configuration),
one can invert equation (2) to determine response functions for the lesion in terms of measurable
quantities:
ow (7)
8En
where Aa are the local principal stretches (i.e., F = diag [>11 , .\ 2 ], which can be calculated from
the motion of a set of closely spaced markers that are measured with video-tracking algorithms
(see Hsu et al., 1995). Hence, information on the specific functional form of the constitutive
relation can be inferred, in principle, directly from experimental data.
Regardless of how the form of w is determined (e.g., directly from data, from microstructural
arguments, or by trial and error), the next step is to determine best-fit values of the material pa-
rameters (e.g., c and ci in equation (1 )), as done in equation (4). In the case of axisymmetry, this
is again straightforward since the principal stress resultants are known in terms of experimen-
tally measurable quantities (equation (6)). Moreover, because of the local nature of Aa and "'a
(and hence Ta), this allows one to explore possible heterogeneous behavior by simply making
measurements in multiple regions. Of course, having separate data in the two principal directions
allows one to explore the possible regional anisotropy (within a restricted group).
Inverse Finite Elements. Most saccular aneurysms are not axisymmetric, however, and as a re-
sult there are no closed-form solutions to infer the form of the strain-energy function directly
2 Histologically more complex lesions or situations requiring viscoelastic or thermoelastic constitutive
relations have not yet been considered.
from experimental data. Indeed, non-axisymmetry complicates even the determination of best-
fit values of the material parameters (e.g., c and Ci in equation (1 )). Perhaps the best approach
to determining the values of the material parameters in this case is the inverse finite element
method. Briefly, one uses a finite element solution to compute the theoretical values to which
experimental values are compared in the nonlinear regression. Whereas stress resultants cannot
be 'measured' in non-axisymmetric lesions, the displacements (i.e., motions of surface markers)
can be measured- note, the displacement vector u = x - X, where x is the deformed location
and X the undeformed location of a material particle. For example, in this case the objective
function e can be written as
L,
m
e= ~ [(xllfem- xllexp) 2 + (x2lfem- x2lexp) 2 + (x3lfem- X3lexp) 2 (8)
k=l
where the subscripts fern and exp denote finite element computed and experimentally measured,
respectively, and (x 1 , x 2, x 3) denotes the location of a particular marker at the m different equi-
librium configurations. As may be imagined, however, the inverse finite element method can be
computationally expensive; it requires a complex nonlinear finite element problem to be solved
for each parameter set considered by the nonlinear regression, which is an iterative process. In
an attempt to address this issue, Seshaiyer et al. (200 1) offered a slightly modified approach: a
sub-domain inverse finite element method. Briefly, this approach builds on the work by Kyriacou
et al. (1997) and others but only requires finite element solutions over a sub-domain D 8 , rather
than over the entire domain D of the physical problem. For example, they considered one such
sub-domain D s that is demarcated by 5 nodes and 4 connecting triangular elements on an inflated
membrane. In this example, there are 3 displacements Ui (or deformed positions Xi) at each of
the 5 global nodes, thus yielding a finite element problem consisting of only 15 equations and
15 unknowns. Yet, if each of these displacements is measured at each distension pressure P,
one could prescribe some of them as 'boundary conditions'. If, for example, we prescribe the 4
outer nodes as displacement boundary conditions, then our finite element problem reduces to the
solution of 3 displacements at the center node at each pressure P, that is, at each equilibrium
configuration m. The objective function is thus based on information at the center node.
Here, we emphasize that this sub-domain method is philosophically consistent with the ap-
proach typically adopted by an experimentalist. In contrast to the theorist who often seeks to
quantify the complete stress and strain fields (i.e., at all positions x), the experimentalist typ-
ically seeks information in selected regions wherein the fields are homogeneous or nearly so.
A prime example is the gage length or central region employed in uniaxial and biaxial testing,
respectively. The sub-domain method exploits such regions. To obtain sufficient information,
therefore, the experimentalist often collects data in the same region but at multiple equilibrium
configurations.
Sub-domain results from a numerical evaluation study - for the axisymmetric inflation of a
Fung-type (1) membrane at multiple distension pressures, which requires the estimation offour
material parameters- revealed that the method is reasonable, provided of course that one has
reasonable initial guesses. As expected, the goodness of the estimates decrease as the experimen-
tal error increases and as the initial guesses are farther from the true values. Of course, central to
any estimation is identification of a good specific form of the relation to start with. In soft tissue
biomechanics, this is often the most challenging step in the formulation. Seshaiyer et al. (2002)
192 J.D. Humphrey
recently reported results for two human aneurysms using the Fung strain-energy function and the
sub-domain method. In comparison to the values obtained by Kyriacou and Humphrey (1996),
based on the data of Scott et al. (1972), the sub-domain determined parameters suggest a much
stiffer and less extensible behavior. Based on a discussion with Professor M. Roach, it appears
that this difference may be due in part to an underestimation of the reference configuration by
Scott et al. (1972), whichwould overestimate the stretches. Regardless, there is clearly a need
for more data on human intracranial saccular aneurysms.
1.4 Conclusions
Rupture of intracranial saccular aneurysms is a devastating event; it continues to be the cause of

significant morbidity and mortality. It has long been thought that the natural history of these le-
sions depends, in part, on the biomechanics. For example, cellular responses to hemodynamically-
induced loads and outright stress-induced failure of the connective tissue network. Fundamental
to diagnosis, prognosis, and treatment, therefore, is a deep understanding of the material prop-
erties of the tissue throughout its enlargement as well as criteria for its rupture. Some data are
available, as are some new experimental, theoretical, and computational methods, yet the need
for additional research is acute.
2 Membrane Theory
As noted in Section 1, the first step in any constitutive formulation is delineation of the general
characteristics of the material behavior. Extensive experience with human saccular aneurysms
suggest that many are thin-walled, with negligible bending stiffness, and that they exhibit non-
linearly elastic or viscoelastic behavior over finite strains. Consequently, a possible theoretical
framework is the nonlinear theory of elastic membranes. This is seen in Section 1, but here let us
consider in more detail some of the salient aspects of membrane mechanics. 3
2.1 Introduction
Membranes are defined differently in biology and mechanics. In biology, the word membrane
implies a thin layer of tissue that covers a surface or separates a space; examples include the cell
membrane, the basement membrane in the arterial wall, and the visceral pericardium that covers
the heart. In mechanics, the word membrane also implies a thin structure, but more specifically
one that offers negligible resistance to bending. That is, the effects of resultant bending moments
and transverse shears are neglected in membranes in comparison to the responses to in-plane
loads. As it turns out, most 'biological membranes' can be modeled mechanically as membranes
in many situations.
The theory of elastic deformations of membranes is a special case within the general theory
of plates and shells; thus, the reader is referred to the classic texts by Green and Zema (1954)
and Kraus (1967). Although membrane theory is important in many traditional problems in en-
gineering (e.g., analysis of pressure vessels and domed structures), the assumption of negligible
3 Much of this is taken from Humphrey (1998).
bending stiffness is particularly appropriate in the finite elasticity of rubber-like (Green and Ad-
kins, 1970) and biological (Humphrey, 1998) membranes. Indeed, Libai and Simmonds (1988)
discriminate between approximations of the full shell equations and what they refer to as 'true'
membranes such as cloth, biological tissues, and thin structures made of rubber-like materials.
As will become evident below, a remarkable consequence of the membrane assumption is that
equilibrium problems can often be solved without explicit specification of a constitutive equation.
That is, stresses can often be found from the deformed geometry and applied loads using statics
alone. Nonetheless, constitutive relations are essential for obtaining complete information on the
material response.
2.2 Constitutive Relations

There are two basic approaches for modeling the hyperelastic behavior of membranes: one can
formally reduce a three-dimensional constitutive relation to a two-dimensional equation, or one
can regard the membrane as a two-dimensional continuum and postulate the existence of a two-
dimensional strain energy function w (no matter how thin, of course, all structures are three-
dimensional; thus, membrane theory is an approximation regardless of approach). We shall adopt
the latter approach herein, but refer the reader to Green and Adkins ( 1970) for a description of the
former. Note, therefore, that w is defined per initial surface area, not volume, and it is assumed
to depend on the in-plane components of the deformation gradient tensor F, or by material frame
indifference, the in-plane components ofU = v'C, C = FT ·ForE= (C- 1)/2.
Following Humphrey et al. (1992), let a 2-D strain energy function w be defined by
w = HW, where w = w(Cu, C22, C12, C21) (9)
and His the undeformed thickness of the membrane. Clearly, we assume that the 1-2 directions
are in-plane and the 3 direction is out-of-plane. Given the general equation for 3-D hyperelastic
behavior (Humphrey, 2002), we have
2 8(w/H)
tij = detFFiAFjB {)CAB , (i,j,A,B = 1,2,3), (10)
l
where the 3-D deformation gradient can be written as
FiA = [F] = [~~~ ~~~ ~

0 0 A3
(11)
with A3 = h/ H, h being the deformed thickness. Together these equations yield

2
htaf3 = - J FarFf3L1-ac ,
aw (a,/3,F,iJ. = 1,2), (12)
2D TL1
where J2n = FuF22- F12F21 is the determinant of the 2-D deformation gradient tensor F.
Assuming that taf3 do not vary significantly through the deformed thickness of the membrane,
we define Cartesian components of the Cauchy stress resultant tensor T (also called tensions or
membrane stresses) as
(13)
194 J.D. Humphrey
It is interesting to observe that for material symmetries wherein the principal directions of
stress and strain coincide (e.g., initial isotropy), and for a principal deformation given by C =
diag [.Xi, .X~] and R = I (where the polar decomposition theorem is F = R · U), then the
principal stress resultants can be written as
(14)
or,
low low
T1 = .X2 o.X1' T2 = .X1 o.X2' (lS)
which is in the form employed by Pipkin (1968) for a w that depends on the principal values of
the 2-D right stretch tensor U.
More importantly here, however, Humphrey et al. (1992) also showed that in cases wherein
the form of w is unknown, the more general equations (12) can be solved for the three (yet
unknown) response functions, namely
(16)
(17)
(18)
where ow I oC12 = ow I oC21 because c is a positive-definite symmetric tensor. Since the right-
hand sides of these three equations are experimentally 'measurable', the values of the three
response functions can thereby be calculated as a function of the deformation directly from
data. That is, plotting values of the response functions versus components of C (or E since
2ow I oC = OW I oE) will reveal characteristic behaviors of the membrane and thus suggest
specific functional forms of the constitutive relations.
2.3 General Axisymmetric Problems
Although the membrane theory is governed by equations that are simpler than those for 3-D finite
elasticity, complete solutions are still difficult to obtain. Indeed, for this reason, most investiga-
tions- experimental, analytical, and numerical- have focused on axisymmetric problems. By
axisymmetric, we mean that the undeformed and deformed configurations can each be described
by generator curves that are revolved about a common axis and the applied loads are independent
of the angle through which the generator curves are rotated.
Kinematics. Figure 1 shows a typical membrane in its deformed configuration, which is denoted
by a generator curve. It is convenient to refer the generator curves to cylindrical coordinates.
Thus, let a material particle at (R, Z) in the undeformed configuration be mapped to (r, z) in the
deformed configuration; (} =e because of the axisymmetry. It will prove very useful, however,
to parameterize each configuration using a single variable, the arc length. Letting S and s denote
/ r2
/
ds
Figure 1. Schema of axisymmetric membrane in cross-section.
arc lengths in the undeformed and deformed configurations, the generator curves can thus be
defined by R = R(S) or Z = Z(S), and similarly r = r(s) or z = z(s).
It is well known from differential geometry that curvature is an important descriptor of curves
and surfaces. In essence, curvature is a measure of the pointwise change in the orientation of
an outward unit normal vector n on the surface of interest. See standard texts on differential
geometry, or Chapter 1 in Kraus (1967), for a more complete description of curvature. Here,
we are interested in the two principal curvatures (i.e., maximum and minimum values) in the
deformed configuration, denoted by "'a' and which will be easy to visualize. From Figure 1,
therefore, note that the first principal curvature, "' 1 , is defined by
dip dip dr
(19)
"' 1 = ds = dr ds '
where from geometry alone we know that
dr
cos ip = ds, (20)
It is also useful to think of the principal curvatures "'a of an axisymmetric surface as 1/ r a, where
r a are radii of orthogonally positioned circles that are each tangent to the surface at the point of
interest; one of these circles will be in the plane containing the outward normal n and the tangent
vector along s. From Figure 1, therefore, we see that sin ip = r fr 2 . Hence, the two principal
curvatures can be written as
1 sin ip
/),2 =- = --. (21)
r2 r
196 J.D. Humphrey
It immediately follows, therefore, that
d
dr (rii2) = li1, (22)
which is one of the so-called Gauss-Codazzi relations; it will prove very useful experimentally
for it reveals that only one principal curvature needs to be measured directly and it will prove
useful analytically in simplifying one of the equilibrium equations.
Although equations (21) are complete descriptors of the principal curvatures in the deformed
configuration, alternate descriptions are useful. For example, it is expedient to exploit the para-
metric representation r = r(s), which when combined with equations (19) and (20) yields
d<p d ( _ 1 (dr)) d 2r/ds 2 (23)

lil = ds = ds cos ds =- y'l- (dr/ds) 2 ·
Likewise, noting that sin 2 <p + cos2 <p = 1, equation (21) 2 can be written as
y'l- (dr/ds) 2
1i2 = (24)
r
These equations allow the principal curvatures to be determined from a single parametric repre-
sentation of the generator curve, which is again very useful experimentally.
To employ these equations, one needs a C 2 function that describes the profile of the deformed
configuration of the membrane. It is not always easy to find such a function (e.g., polynomial)
based on a finite number of experimental measurements. Rather, it is sometimes preferable to
seek piecewise continuous (interpolation) functions over portions of the domain. For this reason,
therefore, it is sometimes useful to parameterize the arc length s in terms of a new variable, say
(,which is defined over a portion of the membrane. For example, we lets= s(() or
(25)
Hence,
dr
(26)
ds
and similarly
d2 r i(ir - rz)
(27)
ds2 - (r2 + z2)2 .
Insertion of equations (26) and (27) into (23) and (24) yields the familiar relationships for the
curvature. The algorithm in Hsu et al. (1994) shows, for example, how cubic spline interpolants
can be used to calculate curvatures from experimental data\ a procedure that was employed and
verified in Hsu et al. ( 1995).
In addition to curvatures, one also needs local measures of deformation, as, for example, the
stretch ratios (i.e., principal components of the right stretch tensor U). It can be shown that the
principal directions of stretch and stress are the meridional (i.e., s) and circumferential (i.e., 0)
4 Note the typographical error in the exponent in the denominator in equation B7 in Hsu eta!. (1994)
directions in axisymmetric membranes. It is clear, therefore, that principal stretch ratios Ao: are
given by
(28)
Note, too, that these stretch ratios could be measured by tracking two markers that are attached to
the surface along a meridian - these markers would demarcate an approximate arc length LlS,
and changes therein, and their midpoint would identify the point (r, z) at which the stretches
would be evaluated. In analytical formulations, it is also useful to recognize that >. 1 can be rep-
resented as
>.1 = J(dr/ds) 2 + (dz/ds)2. (29)
Equilibrium. There are multiple ways to derive the equilibrium equations for an axisymmet-
rically deformed membrane (e.g., see Green and Adkins, 1970; Kraus, 1967). Here, however,
we consider a problem of uniform pressurization only and thus use a simple free body diagram
approach.
Let the principal stress resultants be denoted by To: and the distension pressure by P. Force
balance in the z direction requires
(30)
where the pressure has been multiplied by the projected area. Now, let rb = r a+ Llr and recall by
the Taylor's series expansion for any quantity, say J, that f(xb) = f(xa) + (df /dx)(xb- Xa) +
higher order terms. Hence, equation (30) can be re-written as
:r (271'rT1 sin<p)Llr = 11'P(2rLlr + Llr 2), (31)
whereby, dividing by Llr and neglecting higher order terms, we obtain
! (rT1 sin<p) = Pr. (32)
Similarly, considering a free-body diagram that exposes circumferential tensions, balance of

forces in the circumferential direction yields
-2T2Lls- (2rT1 cos<p)lr., + (2rT1 cos<p)lr& + P(2rflz) = 0, (33)
where the 2 in the second and third terms came from J sin 8d8 from 0 to 71'. Because Llz =
Llr tan <p and Lls = Llr / cos <p, the second equilibrium equation can be written as
1 d
T2-.-- -d (rT1 cos<p) cot<p
sm<p r
= Pr. (34)
Together, equations (33) and (34) relate the principal tensions To: to the applied pressure P at
each position <p; recalling that the change in <p with arc length s is a measure of the curvature,
these two equations can be combined to yield the alternate form
(35)
198 J.D. Humphrey
Moreover, equation (32) can be expanded, and recalling equations (21 ), written in the more
familiar form
(36)
which is sometimes called Laplace's equation. Together, these are the two fundamental governing
equilibrium equations for an axisymmetrically inflated membrane.
Recalling the Gauss-Codazzi relation (22), equations (35) and (36) can be solved in closed
form without explicitly introducing a constitutive equation. That is, substitution of equations (22)
and (35) into (36) yields
(37)
which can be written as

(38)
Multiplying through by r, assuming that Pis a constant (at each equilibrium configuration), and
integrating thereby yields
(39)
where cis an integration constant. For problems where r = 0, c = 0 and the meridional stress
resultant reduces to
p
T1=-. (40)
2~2
Similarly, appealing to equation (36), the circumferential stress resultant is
(41)
See equations (4.30) in Kraus ( 1967), and the related discussion, for a different derivation of
these results. Although simple in appearance, these equations have not found much use in an-
alytical solutions. The primary reason for this is that the curvatures must be calculated from
the deformed configuration via nonlinear differential equations (recall equations (23) and (24)),
which are challenging. As pointed out by Hsu et al. (1994), however, curvatures are experi-
mentally measurable, and thus equations (40) and (41) are extremely powerful results for the
interpretation of membrane inflation tests; see equation ( 6). Indeed, one cannot overemphasize
the beauty of the membrane theory in that it admits equilibrium solutions without the need to
introduce explicitly a constitutive equation. We say explicitly, of course, for it is the material
properties that govern the configuration (i.e., r, ~ 1 and ~2) that the body will assume under the
action of a particular distension pressure P. Equations (40) and (41) implicitly contain informa-
tion on the material properties therefore (e.g., this could be thought of as the pressure-volume or
pressure-radius relation P = P(r) in the case of a spherical membrane).
2.4 Special Equilibrium Solutions
It is useful to record three special cases that follow from equations (40) and (41 ): the inflation
of membranes that are spherical, cylindrical, or elliptical in both their reference and current (de-
formed) configurations. For this to occur under the action of a uniform pressure, certain material
symmetries must exist, as, for example, initial in-plane isotropy in the case of a sphere inflating
into a larger sphere.
The principal curvatures of a sphere are, of course, f\;1 = t1;2 = 1I a, where a is the deformed
radius. Thus, for an inflated spherical membrane, equations (40) and (41) reduce to
Pa
T1 = T2 = 2 , (42)
a classical relation that is derived in introductory mechanics of materials courses (pressure vessel
problems- it is usually derived by considering a free body diagram of a thin-walled hemisphere
subjected to a distension pressure). We re-emphasize, however, that this result is good for all
thin-walled spheres subjected to a uniform distension pressure P. Because it is not restricted to
a particular class of materials, this is an example of a universal result; it is particularly important
in biomechanics, as, for example, in the analysis of the urinary bladder, the sphering of red blood
cells, etc., and as shown in Sections 3 and 4, intracranial saccular aneurysms.
Similarly, one principal curvature in a cylindrical tube is 1I a whereas the other is 0 (the asso-
ciated radius of curvature being infinite). Thus, for a uniformly inflated, closed-end, cylindrical
membrane,
T - Pa T2 = Pa. (43)
1- 2 '
This universal result is likewise important in biomechanics, as, for example, in airway, vascular,
and urinary mechanics, provided, of course, that the membrane theory applies directly or one
simply seeks an averaged result. Finally, the principal curvatures for an elliptical membrane are
well known (e.g., see equations (4.33) in Kraus, 1967). The associated principal stress resultants
for an inflated elliptical membrane, having deformed principal radii of a and b, are
(44)
where
(45)
Although the heart is a thick-walled organ, these relations have provided some insight into ven-
tricular mechanics. Of course, the spherical results are recovered from the elliptical ones when
a = b. Comparison of these results reveals that the spherical membrane suffers the smallest
stress resultants, which is to say that it is an optimal geometry for resisting an inflation pres-
sure P. Note, too, that at the pole of an elliptical membrane, tl; 1 = tl; 2 = bla 2 ; thus, the stress
resultants are equibiaxial. In contrast, at the equator, f\; 1 = alb 2 and f\; 2 = 1la; thus, the cir-
cumferential stress resultant will go compressive at the equator when a 2 > 2b 2 . Since a (true)
membrane cannot support compression, this will lead to wrinkling. Wrinkling is an important
topic in membrane mechanics, but we simply refer the reader to the associated literature (e.g.,
Steigmann (1990) and references therein).
2.5 Closure
Given these basic results from membrane theory, one can begin to explore experimentally, an-
alytically, and computationally a number of problems in aneurysm mechanics. For example, in
200 J.D. Humphrey
Sections 3 and 4 we shall consider the issues of material instability and stress analysis, respec-
tively. The literature on membrane mechanics is much more general, however, and should be
explored.
3 Issues of Stability
A longstanding question with regard to saccular aneurysms has been- How can a structure that
consists largely of collagen, which exhibits such high stiffness and low extensibility, continue to
enlarge and eventually rupture under the action of consistent physiological pressures? It has been
suggested by many that, for example, saccular aneurysms may enlarge or rupture due to mate-
rial instabilities, and in particular due to limit point instabilities or resonance. Briefly, the limit
point instability of concern is a bifurcation in the equilibrium solution for an inflated membrane.
Rubber balloons commonly exhibit a limit point instability, for example, whereby the balloon
will expand rapidly once a critical stretch (or volume) has been reached, even in the presence
of a diminished pressure. Resonance, on the other hand, is a phenomenon that occurs when a
structure is excited at its natural frequency; this can result in violent motions that may cause
the structure to fail. Here, we review briefly some of the reports that implicate these material
instabilities in the natural history of saccular aneurysms and then consider in detail more recent
(nonlinear) analyses that suggest that at least certain sub-classes of aneurysms probably do not
experience these instabilities.
3.1 Limit Point Instabilities
Motivation. In an attempt to address the question of how aneurysms may enlarge and eventually
rupture, Austin et al. (1989) and Akkas (1990) both suggested that saccular aneurysms suffer
limit point instabilities, that is bifurcations in their quasi-static response to an increased disten-
sion pressure. Austin and colleagues based their conclusions on in vitro experiments on a 'model
lesion' that they constructed by gluing a 0.8 mm thick collagen patch onto the center of a 0.175
mm thick elastomeric membrane, which in tum was fixed around its periphery and inflated from
underneath. Because of the use of the elastomeric membrane, it is not surprising that this model
exhibited a limit point instability reminiscent of the universal solution for a neo-Hookean ma-
terial (see below) as well as abundant experimental results on rubber balloons. Akkas, on the
other hand, reported computational results for the inflation of a neo-Hookean membrane model
of a saccular aneurysm, which also exhibited a limit point as expected. Yet, because aneurysms,
like most soft tissues, tend to exhibit an exponential rather than rubber-like behavior, it is clear
that these studies needed to be revisited. This was done by Humphrey and Kyriacou (1996) and
Kyriacou and Humphrey (1996) based on the Fung-type descriptor of the Scott et al. (1972) data
(equation (1) in Section 1).
Inflation of a Rubber Balloon. Before proceeding to the analyses by Humphrey and Kyriacou,
let us first review the classical result for a spherical rubber balloon. Aside from perhaps the sim-
ple uniaxial stretching of a rubber band, the inflation of a rubber balloon is the most common
experience that the general public has through which they obtain some appreciation of the pe-
culiar mechanical behavior exhibited by rubber. For example, it is well known that one should
repeatedly pre-stretch a balloon to make it easier to 'blow it up'; likewise, although it is often
difficult to blow up a balloon in the beginning, it soon becomes easier to do so, although over-
inflation will result in rupture. These three common experiences relate to the Mullin's effect (i.e.,
a cyclic stress induced stress softening), the material-dependent instability of interest here, and
the failure criterion. See Beatty (1987) for a discussion of the stress-softening and Payne (1974)
for a discussion of the mechanisms of rupture.
To examine the limit point, we seek the pressure-stretch relation for the sphere: P = P(A.).
Toward this end, we recall that the equilibrium (Laplace) solution for the uniform inflation of a
spherical membrane (equation (42) in Section 2) is trivial, T 1 = T2 ::::::: T = Pal2. Moreover, it
is easy to show that the associated principal stretch ratios are ).. 1 = ).. 2 = ).. = alA, where a and
A are the deformed and undeformed radii, and thus F = diag [)..,A.]. The constitutive relation
relates the stress resultant and the stretch, as, for example (equation (12) in Section 2)
2
Ta/3 = -1 FasF[3.t1~C
ow (a,,B,S,Ll = 1,2), (46)
20 u E.t1
where F is the 2-D deformation gradient, ho = det F, w is a 2-D strain-energy function, and
C = FT · F is the 2-D right Cauchy-Green tensor.
A simple model for rubber-like behavior is the 3-D neo-Hookean model, defined by W =
c1 Uc- 3), where W is the 3-D strain-energy (w = HW, where His the undeformed thickness),
c 1 is the single material parameter, and Ic = tr C, where C = FT · F is the 3-D right Cauchy-
Green tensor. Assuming incompressibility and that C = diag [C11 , C22 , C33 ], then det C = 1
requires that c33 = 1I (Cu c22). Hence,
(47)
and similarly for T22· Because T11 and T22 are the only non-zero components ofT, we denote
these principal values by T 1 = T2 ::::::: T, the uniform Cauchy stress resultant or tension. Hence,
T =
Pa
2 ---+
4c (1 - 1)
P(A.) = AA. )..6 ' (48)
where c : : : : c1 H has units of force per length. Now, for a limit point to exist, dPI d).. = 0 for
some).. > 1 (membranes cannot support compression). Clearly,
dP _ ~
d).. - A)..2
(!_ _ ) _ ---+ _
)..6 1 - 0 Acr - 7
1/6 ,....,
,...., 1.38309. (49)
That is, every uniformly inflated neo-Hookean spherical membrane, independently of the initial
radius A, initial thickness H, and shear modulus c 1 , will exhibit a local maximum in the pressure-
stretch curve at Acr = 7116 ; because this result is independent of the value of the shear modulus,
it is called a universal result. This finding is also appreciated easily by plotting equation (48)2
as P versus A.. Whereas the equilibria are stable up to the critical value of the stretch, they
202 J.D. Humphrey
are unstable thereafter - that is, the membrane will continue to inflate in the presence of a
decreasing pressure (which is the reason a party balloon becomes easy to inflate after some
critical volume is reached). Clearly then, the results of Austin et al. (1989) and Akkas (1990)
are not surprising. Because the neo-Hookean model is not a good descriptor of the mechanical
behavior of aneurysms, however, there is need to revisit this issue.
Analysis for Aneurysms. Again, we consider an idealized spherical lesion having an undeformed
radius A, a uniform initial thickness H, and subjected to a uniform distension pressure P. Note,
too, that F = diag [.A, .A] and that the equilibrium solution T = Pa/2 still holds since it is a uni-
versal relation. Now, however, let us consider an isotropic Fung-type exponential model, which
was shown by Kyriacou and Humphrey (1996) to fit data from human aneurysms reasonably
well. Hence, let
(50)
where EAB are the principal (in-plane) Green strains and c and ci are material parameters. Sim-
ilar to equation (48), it is easy to see that
(51)
where c and r (= c1 + c3 ) are material parameters. Because
dP _ 2cr Q ( 1 + _A2
d.A - Ae
1 + 2F (A2- 1
)2) >0 for all A > 1, r > 0, (52)
the Fung model does not admit a limit point; this can also be seen by plotting P(.A) versus .A.
Inasmuch as Kyriacou and Humphrey (1996) found (using finite elements) a similar result for
a more general case of the inflation of an axisymmetric lesion, it appears that at least certain
sub-classes of saccular aneurysms probably do not expand or rupture via a limit point instability.
This finding re-emphasizes the importance of basing one's analysis on appropriate constitutive
relations. One caveat, of course, is that although the Fung-exponential model appears to be a
reasonable descriptor of the mechanical behavior of aneurysms, there remains a need for further
investigation for the 'best' descriptor.
3.2 Dynamic Stability
Motivation. Intracranial saccular aneurysms are subjected to pulsatile blood pressures; hence
it is natural to ask if they are dynamically stable. For example, many have reported bruits in
cerebral aneurysms- that is, audible tones at frequencies >:::3400Hz. Whereas Ferguson (1972)
suggested that these bruits result from turbulence within the lesion, others suggest that they indi-
cate that aneurysms are excited at their natural frequency and thus resonate (Jain, 1963; Simkins
and Stehbens, 1973; Sekhar et al., 1988). Resonance implies large wall motions, indeed violent
vibrations, and thus was hypothesized by some as a potential mechanism of lesion enlargement
or rupture. Surprisingly, there have been but a few analyses of the associated elastodynamics of
the wall, most of which are based on classical membrane shell theory and thus linearized strains
and material behavior (e.g., Simkins and Stehbens, 1973; Hung and Botwin, 1975). Because sac-
cular aneurysms exhibit a nonlinear material behavior over finite deformations, there is a need
for a fully nonlinear analysis. Moreover, given that these lesions are sometimes surrounded by
cerebrospinal fluid (CSF), there is a need to determine if the 'added mass effect' due to the sur-
rounding fluid plays an important role in the dynamics of aneurysms. In this section, we briefly
review the only nonlinear elastodynamics analysis to date, that by Shah and Humphrey (1999).
Hemodynamic Findings. First, however, it is important to note the following findings from hemo-
dynamic studies. Experimental and computational results both reveal that flow-induced wall
shear stresses T w are small in all classes of saccular aneurysms studied to date: maximum values
are ,::::; 5 to 13 Pa, which are less than the 40 Pa needed to induce endothelial cell damage and or-
ders of magnitude less than the pressure-induced in-plane wall stresses, which can be 1 to 10 MPa
(Canham and Ferguson, 1985; Kyriacou and Humphrey, 1996). These findings, coupled with ob-
servations that the maximum wall shear stress typically occurs at the neck, not the fundus where
rupture tends to occur, strongly suggest that intra-aneurysmal pressures are the dominant hemo-
dynamic loads governing stress-induced rupture (Sekhar and Heros, 1981; Steiger et al., 1988;
Foutrakis et al., 1994). This is not to say, of course, that wall shear stresses are not important;
they likely signal the endothelium to express various molecules, including growth factors, that
may help regulate the remodeling of intramural collagen. This latter role has not been explored
in detail, though it needs to be.
Measured in vivo and in vitro (in glass and silastic models), it appears that intra-aneurysmal
pressures are similar in magnitude to those in the parent vessel (Ferguson, 1972; Sekhar et al.,
1988) and that they do not vary 'much' with position within the lesion. Point-wise measure-
ments via micro-catheters are difficult, however, and must be supported by detailed computa-
tions. Numerical models based on rigid lesion geometries, which are thus limited, suggest that
intra-aneurysmal pressures vary only slightly (1-2 mmHg) within a lesion, consistent with mea-
sured cerebral pressure gradients of 1.25 mmHg/cm in the cerebral arteries. In summary then, it
appears reasonable to assume (to first order) that saccular aneurysms are loaded primarily by a
uniform, time-varying distension pressure.
Elastodynamic Analysis. Rather than developing a complex dynamic finite element method to
examine whether the pulsatility of the blood pressure significantly affects aneurysmal wall stress,
it is prudent to begin with a simpler analysis. Hence, although no saccular aneurysm exhibits a
perfectly spherical geometry, the Laplace assumption for a spherical geometry (T = Pa/2)
yields reasonable estimates of wall stress for a small class of lesions (Shah et al., 1997). Fol-
lowing Shah and Humphrey (1999), consider a thin-walled, spherical aneurysm of initial ra-
dius A and wall thickness H that is subjected to a time-varying distension pressure Pi(t) and
surrounded by CSF. Because of spherical symmetry, let the 2-D deformation gradient tensor
F = diag[.\(t),.\(t)], where the uniform stretch ratio .\(t) = a(t)/A with a(t) the deformed
radius, and let the lesion exhibit an isotropic, Fung-type behavior (equation (50)).
It can be shown that the three equations of motion for a membrane reduce to a single differ-
ential equation for a pulsating sphere, of the form
(53)
204 J.D. Humphrey
where p is the mass density of the membrane, h ( = H /)... 2 , if incompressible) is the deformed
thickness, Ur (= a- A) is the radial displacement, T = T(>..) is the constitutively determined
wall tension,"' (= 1/ a) is the curvature in the deformed configuration, and trr(ri) and trr(r a) are
the radial stress boundary conditions on the inner and outer surfaces of the membrane. Assuming
a prescribed, time-varying uniform luminal pressure, with trr(ri, t) = -Pi(t), Milnor (1989)
shows that arterial pressures are well described by a Fourier series representation of the form
N
Pi(t) = Pm +L (An cos(nwt) + Bn sin(nwt)), (54)
n=l
where Pm is the mean blood pressure, An and Bn are Fourier coefficients for N harmonics,
and w is the circular frequency. Milnor (1989) further suggests that N = 10 is sufficient to
describe blood pressure in general, with N < 10 in the distal vasculature. Among others, Fergu-
son (1972) reported microcatheter measured intra-aneurysmal blood pressures in humans. Using
these data, it can be shown that specific values of P m and An and Bn, for the first 5 harmonics,
are A1 = -7.13,Bl = 4.64,A2 = -3.08,B2 = -1.18,A3 = -0.130,B3 = -0.564,A4 =
-0.205, B 4 = -0.346, A5 = -0.0662, B5 = -0.120, all in mmHg, where Pm = 65.7 mmHg.
Note that these lower pressures were recorded in supine, anesthetized patients.
The cerebrospinal fluid (CSF) could similarly be assumed to exert a uniform time-varying
pressure on the outer surface of the membrane, that is a reaction to the pressure-induced disten-
sion of the membrane. This is tantamount to treating the CSF as an ideal fluid (i.e., inviscid and
incompressible). In this case, trr(r 0 ) = -P0 (t), where P0 can be determined by solving the
pressure field in the fluid domain. For an ideal fluid, the governing differential equations are the
balance of mass and linear momentum (i.e., Euler) equations. In the absence of body forces, they
can be written as \7 · v = 0 and- \7 P = p 1 a, respectively, where v and a are the CSF velocity
and acceleration and p f is the mass density of the fluid. Although the CSF cannot be assumed to
be inviscid, in general, it is likely reasonable to assume an incompressible Newtonian behavior
(similar to the behavior of blood plasma). In this case, the governing differential equations are
the balance of mass \7 · v = 0 and the Navier-Stokes form of the linear momentum equations
-\7 P + 1-l \7 2 v + p1b = pfa, where 1-l is the viscosity of the CSF. The stress boundary condition
is trr(r 0 ) = -P0 (t) + 2~-LD~~(r 0 ), where Dis the stretching tensor (D = (L + LT)/2, where
L is the velocity gradient tensor). Because the solution for the ideal fluid can be recovered from
that for the Newtonian fluid, we consider the latter here.
Assuming a 1-D flow in the radial direction~ in a spherical domain, mass balance requires
that
(55)
The function g(t) is determined by requiring a material particle on the membrane to have the
same velocity as the adjacent fluid particle. Hence, at~ = a,
(56)
It is easy to show that for a radial flow, the meridional and circumferential Navier-Stokes equa-
tions require that the fluid pressure P = P(~, t) alone. Hence, the only 'non-trivial' equation is
the radial equation
(57)
which can be integrated. Indeed, integrating for~ E [a, oo) yields the pressure exerted on the
outer surface of the membrane by the surrounding CSF:
(58)
The pressure at 'infinity' P 00 can be assumed to be either constant or time-varying (note: by

infinity, of course, we mean a sufficient distance~ such that v~ (~, t) « v~ (a, t) for all t). Finally,
recalling the constitutive relation for a Newtonian fluid, t = -PI + 2pD, the radial stress
boundary condition on the outer wall is
(59)
Together, equations (53), (58) and (59) yield the final (nonlinear) governing differential equation
(60)
where T(.A) is given by equations (46) and (50) and Pi(t) by equation (54). For those familiar
with vibrations in simple mechanical systems, this equation is similar (in concept) to that for a
forced mass-spring-dashpot system, which is described by mx +ex+ kx = j(t), where m, c
and k are the mass, viscosity of the dashpot, and stiffness of the spring, respectively, f (t) is a
forcing function, and the over-dots denote time derivatives. In contrast to this simple ordinary
differential equation, however, equation (60) is highly nonlinear. Moreover, note that the effect
of the surrounding CSF is potentially important: it introduces an 'added mass' effect as well as
a dissipative (i.e., first order derivative) term as well. Equation (60) is subject to initial condi-
tions on).. and d.A/dt. Finally, note that in the limit as the dynamical effects disappear, equation
(60) reduces to the familiar Laplace equation T(.A) = Pa/2, as used by Canham and Fergu-
son (1985) and Humphrey and Kyriacou (1996), among others, to examine the elastostatics of
saccular aneurysms.
This nonlinear ordinary differential equation is solved easily using numerical techniques such
as Runge-Kutta, particularly by transforming the single second order equation into a system of
two first order equations. This can be accomplished by introducing a change in variable. First,
however, it is often very useful to non-dimensionalize a dynamical equation. Hence, let length,
time, and mass scales be defined as
(61)
where A and H are the undeformed radius and wall thickness, respectively, p is the mass density
of the wall, and c is a material parameter having units of tension (as in the Fung exponential
206 J.D. Humphrey
w = c( eQ - 1)). Using these scales, it can be shown that equation (60) has the non-dimensional
form
(~
X
+ bx) x + ~bx 2 + 4m~ + 2f(x)
2 X X
= F(T), (62)
where b = PJA/ pH, m = tJ/(pcH) 112 , f = T jc and F is a non-dimensional forcing function.

Now, introducing the change of variables Yo = x and y 1 = dxjdT, equation (62) can be rewritten
as
F(T)- 3byif2- 4myl/yo- 2f(yo)/Yo
(63)
Ya 2 + byo
To solve this system of equations, one must specify values of the non-dimensional parameters.
Toward that end, Shah and Humphrey ( 1999) used the following values of parameters, which
they suggested define a representative lesion: p = 1050 kg/m 3 ' A = 3 X w- 3 m, H = 27.8 X
w- 6 m, P! = 1000 kg/m3 , fJ = 1.26 X w- 4 Ns/m 2 , PXJ = 3 mmHg, and c = 0.88 N/m,
c1 = c2 = 11.82, and c 3 = 1.18. See the original paper for complete results. Figure 2 shows
calculated responses of the model lesion subjected to a Pi(t) E [80, 160] mmHg, similar to that
reported by Ferguson (1972) (i.e., using equation (54)), and subject to two different sets of initial
conditions: equilibrium at timet = 0, where -\(0) = Ae and d-\jdt(O) = 0.0 at P(O) = 80
mmHg, or a perturbation from such, where -\(0) = 1.002,\e and d-\jdt(O) = 0.1 s- 1 . Note
that the frequency is about 1.5 Hz, or 90 heart beats per minute, and that the small variations
in stretch (,\ E [1.2, 1.24]) reflect the high stiffness of the lesion in the physiologic pressure
range. In particular, Figure 2b shows that the unperturbed initial conditions yielded a periodic
solution as expected (i.e., a closed path in the phase-plane). Figures 2c and 2d reveal that, for
the case of perturbed initial conditions, this periodic solution served as an attractor (i.e., the
oscillations tend to dissipate and the solution returns to the periodic solution), hence the solution
is dynamically stable. Finally, panel A in Figure 2 compares the pressure-stretch response for this
lesion subject to quasi-static pressures and that for the dynamic case. As can be seen the inertial
effects are not significant in this regard, and the solution can be treated as a series of equilibria.
Whether this observation holds for other situations (e.g., different material parameters, different
geometry, different forcing function, etc.) must be examined individually, but based on the results
of Shah and Humphrey ( 1999) it appears that quasi-static analyses are justified for fundamental
frequencies less than 5-10 Hz, which is to say most physiological and laboratory applications.
Finally, it should be noted that in many cases it is prudent to supplement numerical results
with analytical ones, if possible. In the case of nonlinear dynamics, this is typically possible
only by studying the stability about fixed (i.e., equilibrium) points, which is accomplished by
linearizing the governing equations. Note, therefore, that if the linearized system is (asymptoti-
cally) stable, then the associated dynamical system is stable; if the linearized system is (neutrally)
stable, then the linearized solution does not provide any information on the dynamical system,
and one must rely on the numerics; and finally, if the linearized system is unstable about a fixed
point, then the dynamical system is likewise unstable about that point although it could stabilize
about another equilibrium configuration. It can be shown that linearization about a fixed point
Ae = o: results in a linear system of two equations in terms of two unknowns (see Humphrey,
2002). Writing these equations in matrix form, it can be shown that asymptotic stability requires
-200
C)
J
J:
-
E 150
E
w 100
a:
:J
CJ) 50
CJ)
L.U
a: 0
a..
0.95 1.2 1.3
-
..!!.!
20
w
1-
c(
a: 0
J:
(.)
1-
w
a: B
~ -20
1.2 1.22 1.24
STRETCH RATIO
.-..
..!!!
.,....
20
w
~
!:(
a: 0
:I:
0
tua:
~ 1.24
-20 c
----- ..;:.---16.25
----=-: 10 411.0515.6
---~-0.-654.555.2 .
STRETCH RATIO 1·2 O TIME (s)
Figure 2. The response to pulsatile pressurization reveals a closed path in the phase-plane (i.e., periodic
solution) as expected (panel B). The short-lived oscillation about the periodic solution (panel C), given an
initial small perturbation in stretch and stretch rate, reveals that the periodic solution is an attractor, and
thus the solution is stable (note: the multiple, nested, decreasing cylinders are hard to see due to the density
of the data points). Panel A shows that the static and dynamic pressure-stretch response coincide, again
reinforcing that the view that dynamics is not important with respect to the overall mechanics.
208 J.D. Humphrey
that the trace of the associated 2 x 2 matrix be negative and the determinant be positive, where
4m [] _ reQ (4aT(a 2 - 1) 2 + 2a + 2a- 1 )

trace[]=- _ 1 b 2 , (64)
a + a det - a- 1 + ba 2 '
where m > 0, the non-dimensional viscosity, and b > 0, the non-dimensional measure of the
mass density; recall, too, that r > 0 is a non-dimensional material parameter in the isotropic
Fung constitutive relation (cf. equation (51)). Clearly then, if the viscosity is negligible (i.e.,
m = 0), the lesion cannot be asymptotically stable in the small. That is, the surrounding CSF
plays a key role in stabilizing the dynamic response of this class of elastic lesions. Note, too, that
it is easy to show that det[] > 0 for reasonable values of the material parameters and possible
fixed points a. Hence, based on this model, the lesion is stable in the small consistent with the
numerical results for the fully nonlinear problem.
In summary, this simple dynamic analysis of a class of (nearly) spherical saccular aneurysms
suggests that they are dynamically stable both when Pi(t) is a periodic function having a fun-
damental frequency less than 5-10Hz (the non-autonomous system) and when it is a constant
(the autonomous system). It appears reasonable, therefore, to focus on quasi-static stress analy-
ses for insight into the mechanics, a conclusion supported by Steiger (1990) and many others.
Indeed, because saccular aneurysms are typically thin, membranous tissues subject to low fre-
quency pulsatile pressures,intuition has suggested that the inertial effects would be small. It is for
this reason that most papers on the mechanics of saccular aneurysms have employed quasi-static
analyses.
3.3 Conclusions
Whereas earlier suggestions that aneurysm enlargement and rupture may be due to limit point in-
stabilities or resonance were based on analyses appropriate for rubber-like and linearized material
behavior, respectively, simple analyses based on finite strain elasticity and a Fung-exponential
model suggest that certain classes of saccular aneurysms may well be mechanically stable. In-
deed, this suggestion is consistent with clinical reports that most lesions will not rupture. What
then could cause the enlargement of saccular aneurysms? One possible answer lies in the follow-
ing.
Research over the last two decades in vascular biology and mechanics has revealed ubiqui-
tous stress (or strain) mediated growth and remodeling processes. That is, via mechano-sensitive
responses by endothelial, smooth muscle, and fibroblast cell types, blood vessels are able to mod-
ify their geometry and material properties in a multitude of ways. It appears that this growth and
remodeling, independent of the particular gross manifestation, depends primarily on the balance
or imbalance in the production or removal of constituents in different stressed configurations. It is
highly likely, therefore, that the enlargement of aneurysms is also due to growth and remodeling,
and in particular a long-term turnover of intramural collagen. There is, however, a pressing need
to put this speculation on a firm theoretical and experimental foundation. Thus, much remains to
be understood with regard to the mechanisms behind the natural history of intracranial saccular
aneurysms and there is a need for continued interdisciplinary research.
4 Stress Analysis
Although the exact mechanisms by which saccular aneurysms rupture remain unknown, it is
widely accepted that rupture occurs when wall stress exceeds wall strength. It is remarkable,
therefore, that there have been but a few rigorous studies of even the quasi-static response of
the aneurysmal wall to applied loads. Rather, simplified analyses have been based on electrical
analog models of latex balloons (Austin, 1971; Cronin, 1973), or constitutive relations that de-
scribe the behavior of rubber-like (Akkas, 1990) or linear materials (Simkins and Stehbens, 1973;
Hademenous et al., 1994). As is often the case in soft tissue biomechanics, the primary reason
for the lack of detailed stress analyses remains a lack of the experimental data that are needed to
formulate multiaxial constitutive relations that capture the complex material behavior, including
anisotropy and heterogeneity. Here, therefore, let us consider briefly some of the results that are
available in the literature, most of which are based on the behavior of human aneurysms reported
by Scott et al. (1972).
4.1 Analytical Results

Canham and Ferguson (1985) used Laplace's equation- a universal result for inflated spherical
membranes whereby the tension T = Pa/2, where a is the deformed radius- to estimate
a critical diameter de at which a saccular aneurysm may rupture. They assumed that the total
volume of aneurysmal tissue vr (= 47ra 2 h, where h is the deformed thickness) remains constant
at all transmural pressures P (i.e., that these lesions suffer isochoric motions in a given state of
enlargement), and showed that
1/3
d = ( 40'eVT )
(65)
e 7rP '
where 0' e is a critical wall strength. Estimates of vr = 1 mm 3 , 0' e = 10 MPa (note that Scott
et al., 1972, reported a O'e =1-2 MPa and Steiger et al., 1986, reported a O'e =0.5-1.2 MPa in
uniaxial tests), and P = 150 mmHg suggested a de = 8.6 mm, a reasonable value. Limitations
of this approach are the same as those in the work of Humphrey and Kyriacou (1996) and Shah
and Humphrey (1999), however- the assumption of homogeneous and in-plane isotropic tissue
behavior (even though no constitutive relation was employed explicitly), and uniformity of the
calculated stress and strain fields. The latter suggests that each material point is equally likely to
fail, which does not account for the propensity of rupture at the fundus (Stehbens, 1990).
4.2 Finite Element Methods

Unconstrained inflation. Although analyses based on a perfectly spherical geometry offer con-
siderable insight, there is clearly a need for geometrically and materially more realistic analyses.
As a step toward this end, Kyriacou and Humphrey (1996) and Shah et al. (1997) presented
fully nonlinear finite element analyses of a class of idealized axisymmetric saccular aneurysms.
Specifically, they considered uniformly pressurized lesions having an initially uniform wall thick-
ness H, a truncated spherical or elliptical geometry, and a Fung-type constitutive behavior de-
fined by a 2-D strain-energy function w (recall equation (1) in Section 1),
(66)
210 J.D. Humphrey
where Eaf3 are the physical components of the Green strain. Moreover, they imposed a clamped
boundary condition (i.e., the displacement) at the neck. Not having sufficient data to quantify
possible regional variations in material behavior, they considered a range of stress-strain behav-
iors from homogeneous and isotropic to heterogeneous and anisotropic.
Restricting our attention to a class of axisymmetric problems, let the location of a mate-
rial particle in undeformed and deformed configurations be described via cylindrical coordinates
(R, Z) and (r, z), respectively, with() = 8 due to the symmetry. Similarly, letS and s denote
arc lengths in the undeformed and deformed configurations, with S E [0, L]. The eight fields
of interest, therefore, are the deformed coordinates, stress resultants, curvatures, and stretches
(r, z, Ta, Ka, Aa, with a = 1, 2). Although these fields can be determined by solving eight si-
multaneous algebraic and differential equations (e.g., via Runge-Kutta methods; see Humphrey,
1998), which results in a family of similarity solutions, here we seek weak solutions. That is, we
shall employ finite elements. For example, a virtual work statement can be written as
l 0
(8w)dA-lPn·8xda=O---+ 27r /(8w)RdS-21rP J('lj;~: -¢~;)rds=O, (67)
where dA and da are differential surface areas in the undeformed and deformed configurations,
respectively, 8x is a virtual displacement, and n is an outer unit normal vector. Moreover, 'lj;
and¢ are appropriate interpolation functions. As revealed by equation (67), this axisymmetric
finite element formulation reduces the problem to one-dimension as it should. Note, too, that
incompressibility is not enforced constitutively (which simplifies the numerics tremendously),
but it can be enforced kinematically in 'post-processing' if needed. This is an advantage of the
direct membrane approach. Following Fried (1982), undeformed (R, Z) and deformed (r, z)
positions can be approximated via isoparametric interpolation, such as quadratic shape functions
of the form
1 1
-2((1- (), (1 + ()(1- (), 2 ((1 + (), where ( E [-1, 1], (68)
with the center node at ( = 0 and Gauss points for the numerical quadrature at ( = ±../3/3.
Finally, let S = So + ')'(, where So is the global arc length at ( = 0, and dS = ')'d( with ')' real.
The equilibrium equations result in a system of (geometrically and materially) nonlinear equa-
tions in terms of the unknown nodal positions { ri, zi}. These equations can be solved using the
Newton-Raphson method. Typically, 80-160 elements were sufficient to achieve convergence.
Focusing on undeformed lesion geometries defined by truncated ellipses or spheres, the re-
quirement that elements be of the same length, that is 2')', restricts the specification of the ini-
tial nodal positions { Ri, zj}. For spherical geometries this restriction is satisfied easily: cal-
culate the total arc length L = AP, where A is the initial radius and P the subtended an-
gle. Next, determine the nodal locations via RJ = Acos(Pi) and zj = Asin(Pi), where
pi = P- (j- 1)L1P,j = 1, 2, ... , N with N the global number of nodes, and LlP the change
in angle from node to node associated with a prescribed LlS. Of course, the situation is similar
for initially elliptical lesions except Ri =A cos( Pi) and zi = B sin( Pi), where A and Bare
major/minor radii, and pi are more difficult to obtain since LlP is no longer the same for each
pair of nodes. Yet, nodal coordinates can be identified using cubic spline interpolation: divide P
into small equal angles(>=:::; 0.0001 rad) for which coordinates (R, Z) can be calculated as noted
above. The arc length L is then approximated as the sum of all straight segments defined by each
successive (R, Z) pair. Cubic splines can then be used to parameterize the coordinates (R, Z) in
terms of S, which thus identifies each { RJ, ZJ} given prescribed SJ = (j - 1h.
Using the aforementioned results, Kyriacou and Humphrey (1996) and Shah et al. (1997)
defined a baseline isotropic behavior by the following values of the material parameters in the
Fung relation: c = 0.8769 N/m, c 1 = c2 = 11.82, and c3 = 1.18. For anisotropic behavior,
the values of c and c 3 were taken to be the same, but values of c 1 and c2 were modified to
allow the ratio of c 1 to c2 to vary linearly in S from 1 at the fundus to either 3 or 1/3 at the
neck. Kyriacou and Humphrey accomplished this variation in cl/c2 by ensuring that the value
of w(A 1 = 1.18, A2 = 1.18) was the same at the neck as it was in the isotropic case; Shah et al.
further required that w(A 1 = 1.18, A2 = 1.18) was the same at each point. Whereas the former
allows regional variations in both material heterogeneity (as suggested by the data of Steiger et
al., 1986) and material symmetry, the latter maintains a type of homogeneity and thereby isolates
effects of regional variations in symmetry. Finally, note that the value of c 1 / c2 must equal 1 at
the fundus due to axisymmetry, which requires T 1 = T2 and A1 = A2 at that location consistent
with the results of Canham et al. ( 1999) on the histology. Moreover, the prescribed boundary
conditions were zero displacement at the neck (i.e., Ur = 0 and Uz = 0 at z = 0, which enforces
A2 = 1 at z = 0, where u is the displacement) and zero radial displacement at r = 0, the
symmetry axis. These boundary conditions also suppress rigid body motions.
Perhaps the first question that one should address with the finite element method is the ap-
plicability of the Laplace equation (Ta = T = Pa/2, that is uniform, equibiaxial principal
stress resultants). Shah et al. attempted this by first finding the best-fit sphere for the deformed
configuration of various model lesions as calculated by finite elements. That is, fit the deformed
generator curve via (rJ) 2 + (zj- o) 2 = a 2 , where j = 1, ... , N is the number of nodes used
in the simulation, and o and a define the center and radius of the best-fit sphere. Next, calculate
the principal Cauchy stress t, which equals T / h or T A2 / H, where h = H /A 2 is the deformed
thickness, A = a/A the uniform stretch ratio, and A the undeformed radius, This requires the
value for the uniform A associated with T, which can be obtained by inverting the constitutive
relation. Finite element and Laplace results were then compared as a function ofundeformed arc
lengthS.
Figure 3 shows results for lesions having three different initial geometries (i.e., values of
A/ B, the ratio of the initial R and Z major/minor axes) but otherwise the same initial lesion
volume, thickness, isotropic material behavior, quasi-static distension pressure, and boundary
conditions. The undeformed generator curves (dashed lines in panels A to C) reveal that the
prescribed geometry was one half of a complete ellipse or sphere. Panels A to C show that the
initially elliptical or spherical geometry was distorted upon loading (see solid lines) due to the
fixed boundary condition at the neck. Despite equal increments in pressure from 0 to 80 and
then 80 to 160 mmHg, most deformation occurred at lower pressures, as expected of a material
that exhibits an exponential stress-strain behavior. Panels D to F reveal a number of important
observations with regard to the distributions of the principal Cauchy stresses t a: the meridional
stress (solid curve) was higher than the circumferential stress (dashed curve) in lesions when
A/ B 2: 1, but the converse was true when A/ B < 1; the highest multiaxial stresses occurred at
the fundus in the lesions with the highest ratio of A/ B; the highest multiaxial stresses occurred
near S / L ~ 0. 7 (with S / L = 0 at the fundus and 1 at the neck) for lesions with A/ B < 1;
and, as expected, stresses were uniform over the largest domain in lesions having A/ B = 1,
Figure 3. Results for finite element simulations of three lesions having initial geometries defined by A/ B =
0.32, 1.0, 3.12, but otherwise the same initial lumen volume (0.0398 mL), wall thickness (H = 27.8 mm),
and isotropic material properties (c1 = c2 at all S). Results are for two different equilibrium pressures:
80 and 160 mmHg. Undeformed and deformed configurations are given in panels A-Cas dashed and solid
lines. Panels D-F show stress fields in the meridional (solid curves) and circumferential (dashed) direction
as a function of a nondimensional undeformed arc lengthS/ L (with 0 denoting the fundus and 1 the neck);
the dotted lines are the Laplace solution associated with the best-fit sphere to the deformed configuration.
Panels G-1 show similar results for the principal stretches.
that is an initially spherical geometry. Panels G to I show the associated distributions of the
principal stretch ratios Aa. Note that .A 2 = 1 at S / L = 1 as required by boundary conditions.
Finally, the dotted lines in panels D to I show the uniform stress and stretch values predicted by
the Laplace solution based on the best-fit sphere for the deformed configuration. As expected, a
Laplace approximation was best for the initially spherical geometry although one may argue that
a reasonable mean value for Ta(S) was obtained in each case.
Similar simulations were performed for the same three lesions and loading conditions with
the exception that the material properties varied linearly inS from isotropic at the pole (i.e., c 1 =
c2 = 11.8atS/L = O)tocircumferentiallystifferattheneck(i.e.,c1 = 5.93andc2 = 17.79at
S/ L = 1). This resulted in a decreased equibiaxial stress at the fundus, a maximum multiaxial
stress away from the fundus, and an increased meridional stretch near the neck. Increasing the
circumferential stiffness toward the neck when A/ B > 1 thus tended to homogenize the stress
field. Similar results were determined for the same three lesions except for material properties
that vary linearly in S from isotropic at the fundus to meridionally stiffer at the neck (i.e., c 1 =
17.79 and c2 = 5.93 at S/ L = 1). The deformed configurations were similar to those in Figure 3,
and so too for the stress and stretch fields with two exceptions. The maximum stresses increased
at the fundus and the meridional stretch decreased at the neck when A/ B 2 1, and the maximum
values of circumferential stress and stretch (i.e., nearS/ L = 0. 7) increased slightly with respect
to those for A/ B < 1.
Finally, simulations were performed for lesions having three different initial sizes (i.e., lu-
minal volume) but otherwise the same initial spherical shape (i.e., the same undeformed radius,
but truncated at different locations), thickness, isotropic material behavior, quasi-static disten-
sion pressures, and boundary conditions. As expected, the more completely spherical geometry
yielded the most sphere-like behavior: the stress and stretch fields were uniform and equibiaxial
over a large portion of the lesion, the only variations being due to the boundary condition at the
fixed neck. Despite marked differences in size (undeformed and deformed), however, the magni-
tude of the stresses and stretches were nearly the same at the fundus and similar over the entire
domain. This is expected, of course, for it is curvature, not overall size, that controls equilibrium
in inflated axisymmetric aneurysms - this has not yet been appreciated clinically and may be
one reason for the longstanding controversies with regard to the 'critical' size. As expected, the
Laplace approximation was very appropriate for the nearly complete sphere-like lesions but less
so for cap-like lesions.
As a final comment, recall that rupture occurs most often at the fundus. These simulations
revealed that the maximum normal stress may occur at the fundus, in the equatorial region, or
at the neck depending on the geometry, distribution of material properties, etc. In many cases,
however, the maximum multiaxial stresses were found in and near the fundus, with tq,q, = tee.
Recall, therefore, that the maximum shear stress t'q,e = (tq,q, - tee) /2, that is one-half the dif-
ference in the principal stresses. Because t'q,e = 0 at and near the fundus in most cases, it would
seem that failure is unlikely to be due to shear. Indeed, this is consistent with numerous reports
that blood vessels tend to fail in extension (Humphrey, 1995). There is clearly a need for much
greater attention to the formal identification of a rupture-criterion, however.
Contact constraints. Although most saccular aneurysms tend to remain asymptomatic, those un-
ruptured lesions that present with symptoms often do so by pressing on adjacent tissue, including
nerves. A key question, therefore, is whether symptomatic lesions have a greater rupture poten-
tial than asymptomatic lesions. In an attempt to begin to address this question, Seshaiyer and
Humphrey (200 1) presented results for model axisymmetric lesions subjected to a flat, rigid con-
tact at their fundus -this simple case preserves the axisymmetry, thus simplifying the analysis.
The basic finite element model was the same as that employed by Shah et al. (1997) and
discussed above. Note, however, that there are multiple ways to model a contact problem. One
way is to include the contact condition in the weak form via a penalty method. Alternatively,
214 J.D. Humphrey
a.
e Preferred
.s 0.5
N
N Isotropic~-------.
0 0.5 1.5
R, r(mm)
500
'ii
Q.
--- ... .:::::::----.
-,_
.....
X.
400 -,_
0
0
w 300
0::
1-
en 200
>-
:t
0 100
:l ·,_ I
c(
0 '·'
0·
0 0.5
S/L
Figure 4. Lesion defined by B > A and Zr = 0.9. The preferred properties are defined by p = 3 and
(c2/cdmax = 0.25 (see equation (68)). Note that the preferred properties tend to homogenize the stress
field.
one could solve the problem by introducing a 'slack variable' and deriving an analytic expres-
sion for the variational problem. Yet another way is to exploit the fact that when a finite element
node comes into, and stays in, contact with a fixed obstacle it loses a degree (or degrees) of
freedom. For example, instead of satisfying two separate equilibrium equations, only one equi-
librium equation would have to be satisfied at the node along with a constraint condition. This
is equivalent to suitably modifying the boundary conditions as the nodes contact the boundary
during incremental inflations. Seshaiyer and Humphrey adopted the latter approach given the
simplicity of the axisymmetric problem that was considered.
Figure 4 shows two possible cases: a small indentation due to a constraint (panels B, E) and
a more dramatic indentation (panels C, F), which could also be visible on angiography. As it
can be seen, in both cases the constraint is protective. That is, in comparison to the associated
unconstrained lesion (panels A, D), the multiaxial stresses are either less or not changed due to
the constraint. Indeed, Seshaiyer and Humphrey (2001) show that only very small constraints
(i.e., essentially point loads) increase the stress field and thus the potential for rupture.
4.3 Towards a Growth Model

Remarkably, there has been but one report of a computational model for studying the growth
of saccular aneurysms, and unfortunately it is not described in detail. Briefly, Steiger (1990)
considered a class of axisymmetric lesions and stated that 'tissue growth rate was set proportional
to wall stress.' Although there is no discussion of the constitutive or evolution equations, he
reports that 'sausage-shaped and disc-shaped' lesions tended to develop toward a spherical shape
whereas multi-lobed lesions tended to remain complex. He further suggested that localized blebs
may initiate in an attempt to stabilize a localized weakness in the wall.
Here, recall two things. First, non-complicated saccular aneurysms tend to consist largely of
collagen and fibroblasts, the latter of which likely arise in part from a remnant of the adventi-
tia. Moreover, due to the presence of residual stress, the adventitial fibroblasts likely experience
nearly uniform and equibiaxial stresses in the parent vessel, which defines their normal state.
Second, recall that the aforementioned simulations (e.g., for lesions with meridionally or circum-
ferentially stiffer behavior versus an isotropic behavior) revealed a dramatic influence of regional
differences in material symmetry on the distribution of stress in an axisymmetric aneurysm. In
particular, Shah et al. (1997) showed that the stresses tend to be smaller, closer to equibiaxial,
and more homogenized in certain model lesions that become circumferentially stiffer in a linear
fashion from the fundus to the neck. Based on these two observations, one might ask the follow-
ing question: can fibroblasts in an enlarging aneurysm organize newly synthesized collagen such
that the resulting intramural stresses mimic the normal values experienced in the parent vessel?
Mechanically, this question can be rephrased as follows: does an optimal distribution of material
properties exist that would tend to minimize and homogenize the stress distribution?
In an attempt to begin to examine the latter question, Ryan and Humphrey (1999) defined 12
sub-classes of small lesions (maximum dimensions~ 2 mm) via the triplets (A, B, Zr ), where
A and B are major or minor radii and Zr is a 'truncation level' (which defines the location at
which a complete sphere or ellipse is 'cut-off' to obtain geometries such as in Figure 3). Using
finite element simulations, they then determined preferred material properties in terms of two
quantities, (c2/c1)max E [1/11, 11] andp E [1, 6], where
c2 = 1 + [( c2 ) _ 1] ( ~. 1 )p (69)
C1 C1 max N - 1 '
and c2 and c1 are the material parameters in the Fung pseudostrain-energy function, (c 2I c 1)max
is the ratio of these parameters at the neck of the lesion (i.e., at S = L ), i E [1, N] is the finite
element number, and pis a descriptor of how (e.g., linearly or nonlinearly) the material symmetry
varies from the fundus to the neck. For example, (c 2lcdmax = 1 implies isotropy at all 5,
(c2 I cdmax > 1 yields a progressively increased circumferential stiffness toward the neck, and
(c2 I c1 )max < 1 yields a progressively increased meridional stiffness toward the neck. Likewise,
p = 1 requires the symmetry to vary linearly from the fundus to the base (as in Kyriacou and
Humphrey, 1996 and Shah et al., 1997), whereas p > 1 allows nonlinear variations. By preferred,
it is meant that particular combination of (c 2I c 1)max and p that minimizes and homogenizes the
stress field.
216 J.D. Humphrey
Based on literally thousands of simulations, it was found that the multiaxial stresses tend
to be lower and nearly homogeneous in lesions having an initially large neck to height ratio if
p > 1 and (c2/cdmax > 9. For example, Figure 5 compares the different stress distributions for
isotropic (i.e., (c2/c1)max = 1 and p = 1) and the preferred properties. With the exception of the
boundary layer effect (due to the imposed zero displacement boundary condition at the neck),
the stresses are nearly homogeneous for the preferred distribution of properties. Although results
were different for the different geometries (see the original paper), the key finding was consistent
with that of Steiger (1990): small non-complicated lesions (i.e., thin and collagenous, free of
atherosclerosis, fibrin patches, etc.) tended to 'prefer' material properties that allowed them to
become more spherical in shape. For large neck to height ratios this requires that the lesion
expand more in the z direction, which requires less stiffness in the meridional direction; for small
neck to height ratios, this requires that the lesion expand more in the r direction, which requires
less stiffness in the circumferential direction. These findings are teleologically acceptable for the
sphere is the optimal geometry to resist a distension pressure (cf. equation (44) in Section 2). An
unexpected finding, however, was how the lesions preferred to achieve this. For example, lesions
with large neck to height ratios tended to concentrate the anisotropy near the neck (i.e., larger
values of p ). Of course, it is near such a displacement boundary condition that one would expect
large gradients in strain and stress.
Albeit based on idealized models, the finding that intramural stresses in aneurysms could be
returned toward normal values via a preferential deposition of collagen is provocative. Indeed,
given the recent reports that apoptosis (i.e., programmed cell death) and MMP (matrix metallo-
proteinases, which degrade matrix) activity is increased in saccular aneurysms, and so too for the
transcription of type III collagen (the type of collagen that is often formed first in a wound heal-
ing response), it is very reasonable to expect significant stress-mediated growth and remodeling
in aneurysms (Canham eta!., 1999). It is tempting to hypothesize, therefore, that stable lesions
are those that have remodeled in such a way that the stresses are nearly normalized (i.e., home-
ostatic). Complicating the attempt to grow and remodel, however, may be the insidious effects
of atherosclerosis, the activation of platelets, etc. Nonetheless, there is clearly a need to explore
growth and remodeling theories, which must be based on good estimates of the wall shear stress
and pressure fields that likely serve as signals to the endothelial cells and fibroblasts that control
matrix turnover. Moreover, there is a need for better in vivo data on the time-course of changes
in lesion geometry and histopathological data on regional variations in the microstructure.
4.4 Conclusion
In conclusion, intracranial saccular aneurysms likely enlarge due to a stress mediated growth
and remodeling process and they likely rupture when wall stress exceeds wall strength. Hence,
biomechanics and mechanobiology are fundamental to understanding the natural history of these
potentially devastating lesions. Because of the complex material properties, large strains, and
complex geometry, the finite element method is likely the best method for studying the mechanics
of these complex lesions. There is, of course, a pressing need for more complete information
on the mechanical properties. In addition, there is a need for additional finite element studies.
To date, all analyses have been for axisymmetric lesions. Most lesions are non-axisymmetric,
however, thus there is a pressing need for further development of the finite element models,
8 4
A D
0
4 8 0 0.6 1.2
8 £'4
'E B 6 E
.s (/)
(/)
N w
N4
1- g:2 ~·· ...
(/)
I
CD >-
UJ I
()
I :J
0 <l::O
0 4 8 () 0 0.6 1.2
8 4
c F
0 ,___ _ _ _ ____J
4 8 0 0.6 1.2
RADIUS R,r (mm) UNDEFORMED ARC LENGTH (S/L)
Figure 5. Finite element simulation of an anisotropic lesion defined by a neck-to-height ratio greater than
1.0 (~ 4 : 1) coming into contact with a 1 mm long obstacle. Undeformed and deformed configurations
are given in panels A-C as dotted and solid curves, respectively. Associated Cauchy stresses are shown in
panels D-F, respectively, in both the meridional (solid) and circumferential (dotted) direction as a function
of a non-dimensional undeformed arc length S / L, with the fundus at S = 0. The 'pairs' of results are for
P = 80 and 160 mmHg. Results are for the constraint placed at 90% (panel B) and 30% (panel C) of the
maximum displacement achieved by the center node in the unconstrained case. Note the protective effect of
the constraint with regard to wall stress.
218 J.D. Humphrey
which requires improved clinical information on the geometry and applied loads as well. Much
has been learned, yet much remains to be done.
Dedication
This chapter is dedicated to my brother-in-law, Mr. Harry Zinn, who suffered a subarachnoid
hemorrhage due to a ruptured arterio-venous malformation over 4 years ago and remains in a
coma, and to my sister, Mrs. Bonnie Zinn who cares for him at home on a daily basis in a
remarkable way. When tragedy strikes a family member, the importance of research becomes
even clearer. I hope that this chapter on intracranial aneurysms will motivate many to study the
mechanics of neurovascular disease, for the consequences are great.
Acknowledgements
With regard to this chapter on intracranial saccular aneurysms, I am pleased to acknowledge two
colleagues who have contributed much to my understanding- Dr. Daniele Rigamonti, Professor
and Chief of Cerebrovascular Surgery at the Johns Hopkins University and Dr. Peter B. Canham,
Professor and Chair of Medical Biophysics at the University of Western Ontario. Additionally,
I would like to acknowledge the important contributions of a former post-doctoral fellow, Dr.
Padmanabhan Seshaiyer, Assistant Professor of Applied Mathematics at Texas Tech University,
and those of former Ph.D. students Drs. Frank P.K. Hsu, Stelios Kyriacou, and Amit Shah. Fi-
nally, I wish to acknowledge financial support from the Veterans Administration, American Heart
Association, National Science Foundation, and the National Institutes of Health.
References
Akkas, N. (1990). Aneurysms as a biomechanical instability problem. In Mosora F., ed., Biomechanical
Transport Processes. Plenum Press. 303-311.
Austin, G. (1971). Biomathematical model of aneurysm ofthe Circle of Willis: The Duffing equation and
some approximate solutions. Math. Biosci. 11:163-172.
Austin, G.M., Schievink, W. and Williams, R. (1989). Controlled pressure-volume factors in the enlarge-
ment of intracranial saccular aneurysms. Neurosurg. 24:722-730.
Beatty, M.F. ( 1987). Topics in finite elasticity: Hyperelasticity of rubber, elastomers, and biological tissues
-with examples. Appl. Mech. Rev. 40:1699-1734.
Bruno, G., Todor, R., Lewis, I. and Chyatte, D. (1998). Vascular extracellular matrix remodeling in cerebral
aneurysms. J Neurosurg. 89:431-440.
Canham, P.B. and Ferguson, G.G. (1985). A mathematical model for the mechanics of saccular aneurysms.
Neurosurg. 17:291-295.
Canham, P.B., Finlay, H.M., Dixon, J.G. and Ferguson, S. (1991). Layered collagen fabric of cerebral
aneurysms quantitatively assessed by the universal stage and polarized light microscope. A nat. Record
231 :579-592.
Canham, P.B., Whittaker, P., Barwick, S.E. and Schwab, M.E. (1991). Effect on circumferential order of
adventitial collagen in human brain arteries. Can. J Physiol. Pharmacal. 70:296-305.
Canham, P.B., Finlay, H.M. and Tong S.Y. (1996). Stereological analysis of the layered structure of human
intracranial aneurysms. J Microsc. 183:170-180.
Canham, P.B., Finlay, H.M., Kiernan, J.A. and Ferguson, G.G. (1999). Layered structure of saccular
aneurysms assessed by collagen birefringence. Neural Res. 21:618-626.
Chyatte, D., Reilly, J. and Tilson, M.D. (1990). Morphometric analysis of reticular and elastin fibers in the
cerebral arteries of patients with intracranial aneurysms. Neurosurg. 26:939-942.
Cronin, J. (1973). Biomathematical model of aneurysm of the Circle ofWillis: A qualitative analysis ofthe
differential equation of Austin. Math. Biosci. 16:209-225.
de Ia Monte, S.M., Moore, G.W., Monk, M.A. and Hutchins, G.M. (1985). Risk factors for development
and rupture of intracranial berry aneurysms. Am. J Med. 78:957-964.
Ferguson, G.G. (1972). Physical factors in the initiation, growth, and rupture of human intracranial
aneurysms. J Neurosurg. 37:666-677.
Ferguson, G.G. (1972). Direct measurement of mean and pulsatile blood pressure at operation in human
intracranial saccular aneurysms. J Neurosurg. 36:560--563.
Foutrakis, G.N., Yonas, H. and Sclabassi, R.J. (1994). Finite element methods in the simulation and analysis
of intracranial blood flow: Saccular aneurysm formation in curved and bifurcating arteries. Tech. Rep.
6, University of Pittsburgh, Computational Neuroscience.
Fried, I. (1982). Finite element computation of large rubber membrane deformations. Int. J Num. Meth.
Engr. 18:653-660.
Gaetani, P., Tartara, F., Tancioni, F., Rodriguez y Baena, R., Casari, E., Alfano, M. and Grazioli, V. (1997).
Deficiency of total collagen content and of deoxypyridinoline in intracranial aneurysm walls. FEBS
Letters 404:303-306.
Gibbons, G.H. and Dzau, V.J. (1994). The emerging concept of vascular remodeling. Mech. of Disease
330:1431-1438.
Green, A.E. and Zema, W. (1954). Theoretical Elasticity. Oxford: Clarendon Press.
Green, A.E. and Adkins, J.E. (1970). Large Elastic Deformations. Oxford: Clarendon Press.
Hademenos, G.J., Massoud, T., Valentino, D.J., Duckwiler, G. and Vinuela, F. (1994). A nonlinear mathe-
matical model for the development and rupture of intracranial saccular aneurysms. Neural. Res. 16:3 76-
384.
Hegedus, K. (1984). Some observations on reticular fibers in the media ofthe major cerebral arteries. Surg.
Neural. 22:301-307.
Hsu, F.P.K., Schwab, C., Rigamonti, D. and Humphrey, J.D. (1994). Identification of response functions
for nonlinear membranes via axisymmetric inflation tests: Implications for biomechanics. Int. J Solids
Structures 31:3375-3386.
Hsu, F.P.K., Liu, A.M. C., Downs, J., Rigamonti, D. and Humphrey, J.D. (1995). A triplane video-based ex-
perimental system for studying axisym-metrically inflated biomembranes. IEEE Trans. Biomed. Engr.
42:442-450.
Humphrey, J.D., Strumpf, R.K. and Yin, F.C.P. ( 1992). A constitutive theory for biomembranes: Application
to epicardium. ASME J Biomech. Engr. 114:461-466.
Humphrey, J.D. ( 1995). Arterial wall mechanics: Review and directions. Crit. Rev. Biomed. Engr. 23: 1-162.
Humphrey, J.D. and Kyriacou, S.K. (1996). The use of Laplace's equation in aneurysm mechanics. Neural.
Res. 18:204-208.
Humphrey, J.D. (1998). Computer methods in membrane biomechanics. Camp. Meth. Biomech. Biomed.
Engr. 1:171-210.
Humphrey, J.D. (2002) Cardiovascular Solid Mechanics: Cells, Tissues, and Organs. New York: Springer-
Verlag.
Hung, E.J.N. and Botwin, M.R. (1975). Mechanics of rupture of cerebral saccular aneurysms. J Biomech.
8:385-392.
Jain, K.K. (1963). Mechanism of rupture of intracranial saccular aneurysms. Surg. 347-350.
Kosierkiewicz, T.M., Factor, S.M. and Dickson, D.W. (1994). Immunocytochemical studies of atheroscle-
rotic lesions of cerebral berry aneurysms. J Neuropath. Exp. Neurol. 53:399--406.
220 J.D. Humphrey
Kraus. H. (1967). Thin Elastic Shells. New York: Wiley.

Kyriacou, S.K. and Humphrey, J.D. (1996). Influence of size, shape and properties on the mechanics of
axisymmetric saccular aneurysms. J. Biomech. 29: I 015-1022. Erratum ( 1997). 30:761.
Kyriacou, S.K., Schwab, C. and Humphrey, J.D. (1996). Finite element analysis of nonlinear orthotropic
hyperelastic membranes. Camp. Mech. 18:269-278.
Kyriacou, S.K., Shah, A. and Humphrey, J.D. (1997). Inverse finite element characterization of nonlinear
hyperelastic membranes. J. Appl. Mech. 64:257-262.
Langille, B.L. (1993). Remodeling of developing and mature arteries: endothelium, smooth muscle, and
matrix. J. Cardiovasc. Pharmacal. 21 :Sl1-Sl7.
Libai, A. and Simmonds, J.G. (1988). The Nonlinear Theory of Elastic Shells. New York: Academic Press.
MacDonald, D.J., Finlay, H.M. and Canham, P.B. (2002). Directional wall strength in saccular brain
aneurysms from polarized light microscopy. Ann. Biomed. Engr. (submitted)
Milnor, W.R. (1989). Hemodynamics. Baltimore: Williams and Wilkens.
Payne, A.R. (1974). Hysteresis in rubber vulcanizates. J. Polym. Sci. 48:169-195.
Pipkin, A. C. (1968). Integration of an equation in membrane theory. ZAMP 19:818-819.
Ryan, J.M. and Humphrey, J.D. (1999). Finite element based predictions of preferred material symmetries
in saccular aneurysms. Ann. Biomed. Engr. 27:641-647.
Scott, S., Ferguson, G.G., Roach, M.R. (1972). Comparison of the elastic properties of human intracranial
arteries and aneurysms. Can. J. Physiol. Pharmacal. 50:328-332.
Sekhar, L.N., Heros, R.C. ( 1981 ). Origin, growth and rupture of saccular aneurysms: A review. Neurosurg.
8:248-260.
Sekhar, L.N., Sclabassi, R.P., Sun, M., Blue, H.B. and Wasserman, J.F. (1988). Intra-aneurysmal pressure
measurements in experimental saccular aneurysms in dogs. Stroke 19:353-356.
Seshaiyer, P. and Humphrey, J.D. (200 1). On the potentially protective role of contact constraints in saccular
aneurysms. J. Biomech. 34:607-612.
Seshaiyer, P., Shah, A.D., Kyriacou, S.K. and Humphrey, J.D. (2001). Multiaxial mechanical behavior of
human saccular aneurysms. Camp. Meth. Biomech. Biomed. Engr. 4:281-290.
Shah, A.D., Harris, J.L., Kyriacou, S.K. and Humphrey; J.D. (1997). Further roles of geometry and proper-
ties in saccular aneurysm mechanics. Camp. Meth. Biomech. Biomed. En gr. 1: 109-121.
Shah, A.D. and Humphrey, J.D. (1999). Finite strain elastodynamics of saccular aneurysms. J. Biomech.
32:593-599.
Simkins, T.E. and Stehbens, W.E. (1973). Vibrational behavior of arterial aneurysms. Lett. Appl. Engr. Sci.
1:85-100.
Stehbens, W.E. (1990). Pathology and pathogenesis of intracranial berry aneurysms. Neural. Res. 12:29-34.
Steiger, H.J., Aaslid, R., Keller, S. and Reulen, H.J. (1986). Strength, elasticity and viscoelastic properties
of cerebral aneurysms. Heart Vessels 5:41-46.
Steiger, H.J. (1990). Pathophysiology of development and rupture of cerebral aneurysms. Acta Neurochir.
Suppl. 48:1-57.
Steigmann, D.J. (1990). Tension field theory. Proc. R. Soc. Land. A 429:141-173.
Toth, M., Nadasy, G.L., Nyary, 1., Kerenyi, T., Orosz, M., Molnarka, G. and Monos, E. (1998). Sterically
inhomogeneous viscoelastic behavior of human saccular cerebral aneuryms. J. Vase. Res. 35:345-355.
White, J.C. and Sayre, G .P. (1961 ). Experimental destruction ofthe media for the production of intracranial
arterial aneurysms. J. Neurosurg. 18:741-745.
Whittaker, P., Schwab, M.E. and Canham, P.B. (1988). The molecular organization of collagen in saccular
aneurysms assessed by polarized light microscopy. Conn. Tiss. Res. 17:43-54.
Wiebers, D.O., Whisnant, J.P., Sundt, T.M. and O'Fallon, W.M. (1987). The significance of unruptured
intracranial aneurysms. J Neurosurg. 66:23-29.
Wiebers, D.O. et al. (1998) Unruptured intracranial aneurysms-risk of rupture and risks of surgical in-
tervention. International study of unruptured intracranial aneurysms investigators. New Engl. J. Med.
339: 1725-1733.
Remodeling of Arteries in Response to Changes in their
Mechanical Environment
Alexander Rachev
Bulgarian Academy of Sciences, Institute of Mechanics

Acad. G. Bonchev Str., bl. 4 Bulgarian Academy of Sciences 1113 Sofia, Bulgaria
E~mail: rachev@bgcict.acad.bg
Abstract. Arteries are subjected to mechanical forces, which may vary in time. A long-
lasting alteration in pressure and/or blood flow rate causes an adaptive response termed
remodeling. At the macro-level remodeling is manifest as a change in arterial geometry and
a change in mechanical properties of the arterial tissue. A review of the main experimental
findings concerning pressure- and flow-induced remodeling of large arteries is presented.
Theoretical models of volumetric and global growth based on a continuum mechanics ap-
proach are discussed. Some specific biomechanical problems of arterial remodeling associ-
ated with abnormal narrowing of the arterial lumen are considered.
1 Introduction to Growth and Remodeling of Arteries
The physiological function of large arteries is to transport blood from the heart to tissues and
organs and to supply them with nutrition and oxygen. Moreover, arteries transform the highly
pulsatile heart output into a flow of moderate fluctuations serving as an elastic reservoir ('wind-
kessel').
Arteries are almost circular cylindrical tubes oflayered structure (Figure 1). Endothelial cells
(ECs) form a one-cell-thick layer, which together with the underlying membrane is called the in-
tima and is in direct contact with flowing blood. The thickest layer, called the media, is the major
load-bearing structure. It consists of smooth muscle cells (SMCs), collagen fibers, elastin and
ground substance matrix. In most large arteries SMCs are aligned mainly in the circumferential
direction. When appropriately stimulated vascular SMCs contract or relax, resulting in a change
in the arterial diameter and wall thickness and thus affecting the strain and stress distributions
in the vessel wall. Under normal physiological conditions the SMCs are partially contracted and
form the so-called basal muscular tone. The adventitia is the outermost layer and consists mainly
of collagen fibers, ground substance and some fibroblasts.
Arteries are subjected to axial forces due to surrounding tissues; to a periodic transmural
pressure, which varies from diastolic to systolic value; and to flow-induced shear forces applied
at the inner surface due to friction between the arterial wall and the blood. The load varies during
development, maturity, and ageing. Moreover, pressure and blood flow might change transiently
in some physiological states. For instance, physical exercise is accompanied by an increase in
pressure and an increase in cardiac output. Finally, changes in the mechanical environment are
typical for some pathological states. Hypertension is characterized by a chronic increase in pres-
sure. Arteriosclerosis leads to narrowing of the arterial lumen (stenosis) that causes a reduction
222 A. Rachev
Collagen fiber
mooth mu Je cell
Adventitia Elastin
Media
Intima
Endothelial cell
Figure 1. Schematic representation of the structure and composition of an artery (with the kind permission
of H. Achakri).
or even total arrest of blood flow.

Geometrical dimensions and applied loads are the only physical values that can be directly
measured. Describing mechanics of the arterial wall in terms of local measures such as strains
and stresses requires adopting assumptions about geometry, loading and deformation of the ves-
sel. Stress and strain analyses of an artery considered as a thick- and thin-walled tube are given
in Appendices A and B, respectively. In both cases the mean circumferential stress borne by the
arterial wall is
(1)
where P is the mean arterial pressure, ri is the deformed inner radius and h is the deformed wall
thickness.
Assuming the blood to be a Newtonian fluid and to develop Poiseuille flow, the mean shear
stress at the inner surface is
(2)
where Q is the mean blood flow rate, and 17 is the blood viscosity.
The cyclic nature of the cardiac output causes pulsatile waves of pressure, flow, and wall dis-
placements, which propagate through arteries. Under normal physiological conditions the ampli-
tudes of the pressure-induced pulsatile stresses and strains in the arterial wall and the amplitude
of the flow-induced pulsatile shear stress at the arterial wall are much smaller than the stresses
and strains caused by the mean pressure and the mean blood flow rate.
Remodeling of Arteries in Response to Changes in their Mechanical Environment 223
Pressure-induced wall stress and flow-induced shear stress represent the local mechanical en-
vironment sensed by the cellular components of the arterial wall. The SMCs are exposed to the
circumferential wall stress and to the circumferential strain, while ECs are exposed to the wall
shear stress and strains.
Like all tissues whose physiological function is associated with exposure to mechanical
forces, arteries are sensitive to changes in their mechanical environment, i.e. to alterations in
arterial pressure and blood flow. When transmural pressure changes the artery undergoes an in-
stantaneous elastic deformation, which is manifest as a change in the vessel dimensions. This
is the so-called passive response, which is characterized by large deformations and mechanical
non-linearity. Besides its complexity, however, the passive response does not indicate that the
deformation develops in living biological tissue.
If a change in the mechanical environment persists, the passive response might be followed
by a change in the contractile state of the vascular SMCs. This response is called the active re-
sponse and is a typical feature of many tissues and organs whose structure contains muscle cells.
An increase in the circumferential wall stress causes contraction of the SMCs that leads to a
constriction of the artery. This reaction is called the myogenic response or Bayliss effect and is
independent of ECs. It appears even when the artery is denuded of endothelium. On the other
hand, changes in blood flow are sensed by the ECs and may trigger processes that lead to re-
laxation or contraction of SMCs and therefore make the artery constrict or dilate. In general the
active response plays a role as a primary adaptive mechanism tending to keep the flow-induced
shear stress and stress distribution in the wall at their baseline values. This goal is achieved at the
expense of altered contractile states of the smooth muscle cells.
Finally, if the changes in pressure and/or flow are maintained for several days and even weeks,
the artery responds by a gradual change in mass, structure and composition. This long-term re-
sponse of an artery is called remodeling and results in a change in arterial dimensions and a
change in mechanical properties of the arterial tissue. Irreversible changes in the arterial geom-
etry and structure occur also during the period of development and maturation and this process
is termed growth. The growth of arteries is determined by two groups of factors. Firstly, those
of genetic origin, which may include vessel size and topology, the rate of maturity and ageing,
and any inherited tendency towards developing occlusive and degenerative diseases. Secondly,
the growth is modulated by epigenetic factors including age-related changes in the mechanical
environment experienced by the arterial wall. Some authors term the processes of mass change
as growth, property change as remodeling and shape change as morphogenesis (see, for example,
Taber (1995)). In this article the term remodeling is used for the description of any irreversible
alteration in arterial dimensions and mechanical properties caused by mechanical factors.
The main characteristics of the passive, active and remodeling responses of arteries are listed
in Table 1.
At the micro-level remodeling involves interaction of multiple ionic and enzymic pathways,
whose precise nature remains unknown. However, at the macro-level of tissue and organs, the
processes of irreversible deformation and mass supply can be studied efficiently using the exper-
imental and theoretical approaches of continuum mechanics.
The first studies of the influence of the mechanical environment on living tissues date back to
the nineteenth century. In 1869 Wolff formulated his famous law, which states that there exists a
strong interrelation between the structure of a bone and the stress field caused by the loading to
which the bone is subjected. Later Roux (1885) introduced the concept of the functional adapta-
224 A. Rachev
Passive response Change in shape (radius, thickness). Change in orientation of the

No change in mass. structural components.
Active response Change in shape without change in Smooth muscle cells contraction/
mass. relaxation; change in ionic
state of the cells; rearrangement
of contractile proteins.
Remodeling Change in geometrical dimensions Growth, division, cell loss,
that might be accompanied by a migration, change in size, shape
change in mass. and orientation of cells, synthesis
Change in structure and composition and degradation of extracellular
of arterial wall. Change in mechanical matrix.
properties of the wall material.
Table 1. Characteristic of arterial response to changes in mechanical environment.
tion of bone to applied load. For soft tissues, in particular blood vessels, Thoma (1893), on the
basis of studies of blood vessels in the embryonic chick, suggested that mechanical forces play
an important role in the modification of the vascular structure. However, intensive biomechanical
investigations on arterial remodeling were not carried out until recent years. The phenomena of
growth and remodeling in bones and soft tissues being related to stress fields were discussed by
Fung (1990) in his book Biomechanics. Motion, Flow, Stress, and Growth. In a comprehensive
review article Taber ( 1995) discussed the biomechanics of growth, remodeling and morphogene-
sis in living systems. Analysis of experimental and theoretical studies available in the literature,
aspects of mathematical modeling in terms of continuum mechanics and a number of novel re-
sults are given in the monograph Mechanics ofgrowth and morphogenesis (Stein (2000)).
The objective of this chapter is to provide the reader with general introductory informa-
tion and trends in experimental and theoretical investigations of the remodeling of arteries. The
chapter focuses on macro-level effects that follow changes in mean pressure, blood flow or local
disturbances in the mechanical environment of an artery. A number of studies in which the author
has been involved are used to illustrate mathematical modeling in the context of solid mechanics.
2 A Brief Review of Experimental Investigations of Arterial Wall

Remodeling
In general, experimental biomechanical investigations aim to answer what-type questions, such

what are the effects of certain mechanical factors on the function and/or structure of the organ
under study. In particular, experimental studies of arterial remodeling are devoted to the follow-
ing questions: i) How do the geometry and mechanical properties of arteries change in response
to alteration in the mechanical environment? ii) Which mechanical parameters might be associ-
ated with the remodeling as triggers or driving stimuli of the remodeling process?
Three main approaches have been used to study experimentally the effects of changes in the
mechanical environment of arteries: i) in vivo investigations; ii) investigations in organ culture
systems; and iii) investigations in cell culture.
2.1 In vivo studies
In vivo investigations include animal experiments realizing controlled changes in the mechanical
environment of arteries and data analysis on the basis of appropriate mathematical models for
the mechanical response of the arterial wall under normal and altered conditions.
The true remodeling of the arteries can be assessed when geometrical dimensions are mea-
sured and compared under condition of zero-stress to eliminate the contribution of the passive
elastic deformation. It has been accepted that the zero-stress configuration appears when a ring
segment of an artery free of external loads is cut radially. Then the artery springs open and the
cross-section takes the form close to a circular sector characterized by the opening angle (see
Appendix A, Figure 27). This demonstrates the existence of residual strains and stresses in the
arterial wall in the state of no load. Because arteries exhibit basal tone that might affect the
zero-stress configuration (Matsumoto et al. (1996) ), an accurate estimation of the effects of re-
modeling on arterial geometry requires giving an account of the changes in the vasoactive state.
Comparison of the arterial dimensions has to be carried out under equivalent tone conditions, for
instance at the state of maximally relaxed smooth muscle cells.
When geometrical dimensions are measured and compared under physiological conditions
this allows evaluation of the observed changes in the arterial response. They result not only from
remodeling but include also the effects of the altered mechanical environment on the passive de-
formation and muscular tone.
In this section we give a brief review of some results of in vivo studies of arterial remodeling.
Remodeling in response to sustained changes in blood flow. Compensatory enlargement

of arteries in response to increased flow has been demonstrated in animal models and human
vessels. Kamiya and Togawa (1980) showed that an arterio-venous fistula between the canine
common carotid artery and the external jugular vein causes increased flow rate in one portion of
the artery and decreased flow rate in another. The intervention creates a marked alteration in the
flow-induced shear stress at the intima. Over time, however, the deformed arterial radius varied
and tended to restore the normal baseline levels of mean shear stress of about 1.5 Pa. Similar
results have been observed by other authors (see, for example, Langille et al. (1989), Brownlee
and Langille (1991) and Langille (1995)). Figure 2 shows the decrease in diameter of the left
common carotid artery of mature rabbits after left external carotid ligation, which reduces the
blood flow by 70%. Compensatory enlargement in response to increased flow was also recorded
during normal development and hypertension.
Although the blood flow rate changes throughout the arterial system, arteries adjust their
diameter so as to maintain a constant flow-induced shear stress. The uniform shear stress distri-
bution has been recorded experimentally in matured animals (see, for example, Hutchins et al.
( 197 6) and Zamir ( 1979)) and during embryonic development (Taber et al. (200 1)). Theoretically,
the uniform shear stress has been derived from the Poiseuille law (eq. (2)) and an optimization
principle known as Murray's law, which states that the total energy required to drive the blood
and consumed by the vessel metabolism is a minimum (see Fung (1996)).
More detailed analysis of the arterial response to an increase of flow rate shows that it in-
volves two successive phases. Firstly, an acute increase in arterial lumen occurs, resulting from
the temporary dilatation of the artery. The process is mediated by the endothelium through the
release of substances such as endothelium-derived relaxing factor (EDRF), the principal com-
226 A. Rachev
e right carotid
o left carotid
2 3 4
Weeks
Figure 2. Diameter of left and right common carotid arteries after left external carotid ligation (Langille et
a!. (1989) ).
ponent of which, NO, causes relaxation of medial smooth muscle cells. Vasomotor response is
followed by a long-term reconstruction of the media due to proliferation and migration of the
smooth muscle cells in such a way that the undeformed lumen of the vessel increases. Arterial
enlargement causes an increase in wall tension and thereby increases the average circumferential
stress. A compensatory thickening of the arterial wall was observed experimentally, which seems
to restore the normal values of the wall stress (see, for example, Masuda et al. (1989)).
Reduced flow elicits a different response. Again, it comprises two successive processes, but,
in general, remodeling does not follow a time course that is simply the reverse of that observed
under increased flow conditions. At first, the decrease in wall shear stress sensed by the endothe-
lium evokes a smooth muscle contraction and shrinkage of the vessel. Reductions in the release
of EDRF or changes in concentration of specific vasoconstrictors are candidates as mediators of
that process. Sustained decrease in flow provokes further processes, which occur mainly in the
intima and result in wall thickening. Normally, remodeling is a self-limiting process and leads
to a restoration of the baseline value of the wall shear stress. Smooth muscle cells migrate into
the intima and produce matrix proteins, resulting in a structure that resembles the underlying
media. This new tissue is referred to as fibrocellular intimal hypertrophy (Glagov et al. ( 1993)).
Sometimes the remodeling process may not lead to a configuration that maintains the normal
level of wall shear stress. The proliferation process may continue and lead to the formation of
a stenosis or obliteration of the lumen. The grown tissue is matrix-free, poorly organized and
forms so-called intimal hyperplasia (Glagov et al. (1993)). The exact mechanisms involved in
the different modes of arterial wall remodeling in response to altered blood flow remain unclear.
Changes in flow also cause remodeling of endothelial cells in the intima. This is characterized
by loss of cells due to decreased flow or proliferation due to increased flow. The net effect is an
adaptation that maintains endothelial cell surface density despite the changes in arterial diameter
resulting from remodeling of the media (Langille (1995)).
When an artery is denuded of intima, i.e. the endothelial cells are removed, a change in flow
does not cause either acute vasomotor response or wall remodeling (Masuda et al. (1989) and
Langille (1993)). Hence the shear deformation ofthe endothelial cells appears to be the first in
the chain of events that cause a change in smooth muscle tone, resulting in further remodeling of
the wall.
The mechanisms by which arteries adapt to chronic blood flow alterations are different in
young and mature animals. It was found that the carotid arteries of adult rabbits subjected to re-
duced blood flow exhibited decreased internal diameter although no significant changes in vessel
mass and wall constituents were observed (Langille (1993) and Langille (1995)). Similarly, it has
been shown that remodeling elicited by elevated blood flow ultimately produces a vessel with ma-
jor properties similar to a control artery (Langille and O'Donnell (1986) and Fath et al. (1998)).
As for young animals, remodeling results not only in a change in geometrical dimensions but it
also affects wall structure and composition, disturbing the normal process of development and
maturity (Langille et al. (1989)).
Recently a comprehensive study on the flow-induced remodeling of rat aorta was performed
by Hayashi (2000). The authors studied not only the effects on the arterial geometry but also
recorded the changes in the vasoactive response to pharmacological stimulation. Hayashi et al.
showed that after 8 weeks there are no significant differences in the active circumferential stress
calculated at the same circumferential strain among arteries exposed to low flow, high flow, and
normal flow conditions.
Remodeling in response to sustained hypertension. Hypertension is a state of persistent

increase in blood pressure. Though a slow elevation of pressure is a normal tendency accom-
panying ageing, a more severe increase in pressure is one of the major risk factors associated
with the development of many cardiovascular diseases. Liu and Fung (1989) and Fung and Liu
(1989) have studied the relationship between hypertension, hypertrophy, and opening angle of
zero-stress state of arteries following aortic constriction of rat aorta. Aortic banding by a metal
clip of the abdominal aorta causes a persistent increase in blood pressure in the aorta. The au-
thors reported that the inner arterial radius remains practically constant. Because blood flow was
not significantly changed this means that the flow-induced shear at the endothelium maintains it
baseline value. The wall thickness increased rapidly in the first few days after the onset of hyper-
tension and then gradually attains a new homeostatic value (Figure 3a).
Calculated mean circumferential stress at the homeostatic state, using eq. ( 1), is equal to
the stress of the normotensive aorta, indicating that the vessel thickens to restore the mean
circumferential stress. Similar finding have been reported by Vaishnav et al. (1990) and by
Matsumoto and Hayashi ( 1994 ), who used renal constriction to cause sustained hypertension.
The time course of changes in the opening angle exhibits a biphasic pattern (Figure 3b ). The
fast increase in opening angle is followed by a slow decrease to an asymptotic value. In general,
changes in the opening angle combined with the changes in wall thickness tend to restore the
circumferential stress distribution in the vessel wall to control levels (Matsumoto and Hayashi
(1996)).
As for the mechanical properties of the wall material estimated on the basis of the incremen-
tal elastic modulus at the in situ loading conditions, adaptation is much slower. After a relatively
longer period (16 weeks) the modulus of hypertensive rats becomes almost equal to the modulus
of the normotensive ones. This means that mechanical adaptation tends to restore the normal
228 A. Rachev
(a) (b)
0.5 D D 200 i
H/R
D
.o • D
Oil
•
0.4 D 150
• s"' D
"'
"51
0.3 § 100
OJ)
D
·§
• """'
0
•
0.2 50
•
0.1 I I I
0
_ , _ _1
I I I I
0 10 20 30 40 50 0 10 20 30 40 50
Number of Days Postsurgery Number of Days Postsurgery
Figure 3. Time course of (a) wall thickness mean radius ratio and (b) opening angle along the aorta (Liu
and Fung (1989)).
arterial function under induced hypertension (Berry and Greenwald (1976) and Matsumoto and
Hayashi (1994)).
Matsumoto and Hayashi (1996) investigated the histology of the aortic wall of control and
hypertensive rats. They found that the thickening, which occurs mainly in the media, is due to
the smooth muscle hypertrophy and increase in ground substances produced by the SMCs. The
authors showed that these effects are most pronounced in the inner lamellar units of the media.
Because elevated pressure causes a greater increase in the circumferential stress at the inner por-
tion of the wall thickness (see Appendix A) these findings support the hypothesis that remodeling
is induced and 'driven' by the wall stress. Moreover, the non-uniform hypertrophy and produc-
tion of ground substances induced by higher stress at the inner portion of the media can explain
the dynamics of the opening angle following sustained hypertension.
Changes in the active response that accompany pressure-induced remodeling were studied
recently in (Fridez et al. (200la)). Sustained hypertension was caused by ligation of the aorta
be-tween the two kidneys. The results obtained showed that the capacity of the vascular SMCs
to develop maximal active stress is not altered in hypertension. The basal tone increases rapidly
in the acute hypertension phase (2 to 8 days post surgery) and drops back towards control values
at 56 days after onset of hypertension. The myogenic response shifts to lower strains in the acute
hypertension phase to restore itself back to control levels at 56 days post surgery. As found by
other authors wall thickness increases over time; however, change in the vascular tone precedes
the geometrical change. These findings showed that the fast change in the tone serves as a pri-
mary adaptive mechanism that tends to restore the baseline strain and stress distribution in the
aortic wall in response to a change in blood pressure.
Finally, some recent in vivo investigations focus on the combined effects of changes in pres-
sure and flow on arterial remodeling. Hayashi et al. (200 l) reported results on the response of ar-
terial wall to the combinations of hypertension and altered blood flow. These authors showed that
the arterial wall thickens to restore the baseline value of wall stress regardless of the magnitude
of blood flow. Pressure-diameter relationships were recorded in vitro under normal conditions
--=e
OJ)
140
120
Air Filters
e
._., 100
Q,l
"'"'= 80
I.
Q,l
I.
~ 60
0.0 Time (sec) 1.0
Reservoir
Clamps
Pump
Air Filters
Chamber
Figure 4. Schematic representation of an organ culture system (Han and Ku (2001 )).
(Krebs-Ringer solution), under active conditions administered with norepinephrine, and under
passive conditions administered with papaverine. The results obtained showed that an increase
in pressure enhances the vasomotor response.
Summarizing, in vivo experimental studies have shown that arteries predominately change
their wall thickness in response to changes in pressure, while blood flow alterations mainly
affect the vessel diameter. Remodeling is directed to restoring the baseline distribution of the
circumferential tensile stress in the media and flow-induced shear stress at the intima.
2.2 Investigations in Organ and Cell Culture Systems
Though in vivo investigations keep the artery under conditions that are close to physiological
conditions, there are difficulties in precise and continuous controlling and monitoring of the me-
chanical environment and remodeling outputs. Moreover, other factors such as nervous stimuli
and the local hormonal and metabolic environment might affect smooth muscle cell activity and
wall remodeling. To focus solely on the effects of mechanical environment organ culture systems
were used. They provide conditions supporting arterial metabolism and maintaining the arterial
function for a period of several days up to two weeks (Matsumoto et al. (1999), Waliszewski et
al. (1999) and Han and Ku (2001)). A schematic representation of an organ culture system is
230 A. Rachev
shown in Figure 4. It consists of a vessel chamber and a perfusion loop.

The advantage of the culture system is that it allows the vessel to be kept in a well-defined
chemical and nutrient environment and to make it possible to eliminate non-mechanical factors.
The system allows independent control of pressure and flow magnitudes and frequencies and to
evaluate better their contribution to remodeling outputs. Because remodeling in response to alter-
ations in flow takes more time than the time period during which current organ culture systems
can maintain the smooth muscle cell viability, most of the investigations focus on the effects of
altered pressure conditions.
Waliszewski et al. (1999) observed an increase in the opening angle after perfusion of calf
carotids at 150 mmHg for 24 hours compared to the controls, perfused at 100 mmHg. This result
is in agreement with the findings of Fung and Liu and confirms the conclusion that in the early
period following the elevation of pressure arterial remodeling is more pronounced in the inner
portion of the arterial wall. By contrast, the opening angle decreases when the artery is perfused
at 75mmHg.
Mechanical and dimensional adaptation of rabbit carotid artery in organ culture was stud-
ied by Matsumoto et al. (1999). After the rabbit carotid artery has been cultured for 6 days at
zero mmHg (hypotension), at 80 mmHg (normotension) and at 160 mmHg (hypertension) it was
found that the inner diameter was not significantly different among three groups. Wall thickness
of hypertensive arteries increases while wall thickness of hypotensive arteries decreases com-
pared to normotensive vessels. The effects of increased pressure on geometrical dimensions are
similar to those observed in vivo (Matsumoto and Hayashi ( 1994) and Matsumoto and Hayashi
(1996)). No significant difference was found in the pressure-diameter relationship between hy-
pertensive and normotensive groups, while arteries cultured under hypotensive conditions are
less distensible (Matsumoto et al. (1999)).
Geometrical changes of porcine carotid arteries cultured with pulsatile flow to maintain a
physiological mean wall shear stress and under hypertensive (200 ± 30 mmHg) or normotensive
(100 ± 20 mmHg) pressure conditions, were studied by Han and Ku (200 1). They found that
the outer diameter of the hypertensive arteries continually increases over time, which most likely
shows that the artery increases its thickness but keeps unchanged the inner diameter to maintain
the baseline value of the wall shear and to adapt the medial tensile stress. These observations
are again consistent with data from in vivo experiments (Fung and Liu ( 1989) and Matsumoto
and Hayashi (1996)). Diameter response evaluated from pressure-diameter curves recorded be-
fore and after administration of norepinephrine has shown that hypertensive arteries manifested
a stronger contractile response compared to controls as observed in vivo.
Han et al. (2000) reported data for remodeling response of arteries to axial stretch in organ
culture. They found that axial mechanical stimuli increase smooth muscle cell proliferation and
promote longitudinal growth.
The results from organ culture studies have shown that pressure and flow conditions are ma-
jor determinants of the geometrical dimensions, mechanical properties and the active response
of arteries. A limitation for artery organ culture is still the short period of artery viability. A chal-
lenging task is to study the effects of pulsatile pressure and flow on arterial remodeling.
Investigations in organ culture provide important information for controlling the mechanical
environment in bioreactors for manufacturing functional tissue engineered vascular grafts. Con-
structs composed of biodegradable scaffold and seeded endothelial and smooth muscle cells are
subjected to changes in pressure and flow to promote neo-artery formation and to produce an
arterial substitute of appropriate mechanical properties and sufficient strength.

Finally, investigations of cell culture are addressed to study the events that occur when smooth
muscle cells or endothelial cells are subjected to flow-induced shear or to constant or cyclic
strains. It was found that endothelial cells and smooth muscle cells change their shape, orien-
tation, proliferation, release of vasoactive substances and matrix protein secretion (see, for ex-
ample, Leung et al. (1976), Nerem (1993) and Kamiya and Ando (1996)). Though cell culture
is characterized by precise control over mechanical environment, this approach disregards the
interactions among cells and the influence of extracellular environment that exist under physio-
logical in vivo conditions and in organ culture systems. Therefore, the results obtained from cell
culture studies do not allow quantitative prediction of the geometrical and mechanical outputs of
remodeling of native arteries in response to changes in their mechanical environment.
In conclusion, the experimental investigations have shown that arteries remodel their geomet-
rical dimensions to maintain the flow-induced shear stress at the intima and the pressure-induced
wall tensile stress in the media. Remodeling of the mechanical properties is a much slower pro-
cess and might aim at restoring the normal level of the arterial function as an elastic reservoir.
In general, remodeling represents a locally controlled adaptive response tending to cope with
changes in the mechanical environment. Remodeling is preceded by a change in the smooth
muscle activity that occurs as a primary adaptive response to the altered mechanical environ-
ment.
3 Mathematical Models of Remodeling
Experimental evidence that the arterial response to changes in pressure and flow is a local phe-
nomenon, which can be described in terms of mechanical quantities such as strains and stresses,
suggests developing mathematical models based on continuum mechanics. The results of these
models aim to predict the outputs of geometrical and mechanical remodeling caused by the
changes in the mechanical environment. The information is important because the dimensions
and mechanical properties determine the arterial function of distributing and transporting the
blood by means of a flow of moderate pulsations.
The general idea of modeling is to substitute the real object by an abstract model, which
reflects the main characteristics of a certain class of objects under study. The results obtained
as model predictions are claimed to be valid for the real object. In contrast to the experimental
studies that answer what happens, the models aim to answer why does a certain event occur.
Comparison of theoretical predictions and experimental findings justifies the acceptance or re-
jection of the model hypotheses introduced for mechanical quantities that drive and govern the
adaptation process. If no appropriate experiments exist, modeling can suggest the kinds of new
experiments that are needed and methodology for analyzing the data. It is expected that the results
obtained from continuum mechanics models may advance the level of understanding of the role
that mechanical factors play during normal arterial development and maturity, and might help
to reveal the mechanical aspects of the genesis and progression of certain vascular pathologies.
Additionally, results from model studies could promote the development of therapeutic interven-
tions with the aim of restoring the mechanical loads on arteries to normal levels.
Finally, knowledge of mechanisms underlying arterial wall remodeling can be used to design
a favorable mechanical environment for tissue engineered arteries in bioreactors. Although the
232 A. Rachev
results of experimental studies can give insights into the directions that an optimal mechanical
conditioning must follow, these results have a limited predictive value. In fact, the vascular cells,
and the smooth muscle cells in particular, sense and respond not to pressure and flow rate but to
local stresses, which cannot be measured but are to be calculated using an appropriate mathemat-
ical model.
Two main theoretical approaches are used to study remodeling of arteries in response to
changes in their mechanical environment. Both address the kinematics of the remodeling process
and differ in the level at which the geometrical changes are described. The models based on the
volumetric growth consider an artery as a collection of growing differential elements that change
their zero-stress state. Alternatively, the models of global growth focus on the description of
the kinematics of the zero-stress configuration of the vessel as a whole. In this context the term
growth will be used to describe changes in mass that might not be associated with the process of
biological development and maturity.
3.1 Volumetric Growth
General three-dimensional theory for finite volumetric growth has been developed by
Rodriguez et al. (1994) and was used to describe remodeling of arteries by Taber ( 1998b) and
Taber and Eggers (1996)). Following Rodriguez et al. (1994) and Taber (1998b), in this section
we first present briefly the general theory and then its application to the remodeling of arteries to
changes in mean arterial pressure.
Consider an elastic body that at time t 0 has the stress-free unloaded configuration B (t 0 )
(Figure 5). Applying external loads it instantaneously undergoes a finite deformation and takes
the configuration b( t 0 ). The kinematics of the deformation process is described by the defor-
a
mation gradient tensor F 0 = ax( to) I X that transforms the differential position vector dX in
B 0 into differential position vector dx( t 0 ) in b( t 0 ) If the body is made of growing continuum,
the deformed configuration varies over time despite the constancy of the applied load and at the
moment t = t 1 becomes b( t 1 ). This process can be termed as the observed growth. The mapping
of B(t 0 ) into b(h) is described by the gradient tensor F = ax(ti)IaX. The deformed config-
uration b( t 1 ) is a result not only of addition and/or removal of volume, i.e. true growth, but is
also affected by the change in the elastic deformation. If the loads are removed at time t = t1
the body will take a configuration B" (h), which is different from B (to) and might not be in a
stress-free state but contains residual strains and stresses. Thus, the load-free configuration at the
moment t = t 1 is also a result of growth and deformation.
To describe the kinematics of the observed growth on the basis of the true growth and de-
formation process imagine that the body is divided into infinitesimal elements. Each element un-
dergoes finite volumetric growth. This grown stress-free configuration is denoted by B' (h) (Fig-
ure 5) and is the result only of addition/removal of volume to/from each element. The transforma-
a
tion of a differential position vector is defined by the growth gradient tensor F g = ax( tl) I X,
where x( h) is the position vector in B' (t 1 ) of a point having a position vector X in B (to). This
process cannot be isochoric and the volume of a differential element may increase (det F g > 1)
or decrease (detF g < 1). The grown elements in B' (t 1 ) may not fit together. If the growth does
not destroy the continuity of the material, reassembly of the elements into configuration B" (t1)
requires deformation that gives rise to a stress field called the residual stress. The mapping of
B' (ti) into B" (t 1 ) is described by the elastic deformation gradient tensor F el· Finally, after
B(tn)
b(to)
Load
.......
Ob erved growth
Load
I
I
I
/ F•l
Unload ,,
--- ,'
Figure 5. Schematic representation of configurations resulting from deformation and volumetric growth
(modified from Rodriguez et al. (1994) and Taber (1998)).
applying the external loads, the configuration B"(t1 ) transforms into b(t 1 ) and the mapping is
described by the deformation gradient tensor F e2.
The total deformation that the grown body undergoes from its zero-stress state B' (h) to the
deformed state b( tl) at time t = t 1 is described by the total elastic deformation gradient tensor
Fe = F e2 F el· Finally, the transformation that an infinitesimal volume undergoes at an arbitrary
point of the reference configuration B(to) due to the growth and applied forces is described by
the gradient tensor F = Fe F g.
Determining Fe and F g requires solving a boundary-value problem for the coupled deforma-
tion and growth processes. Following the theory of finite deformations the corresponding Green
strain tensor is e = (F~Fe- I)/2, where I is the identity tensor and the superscript T denotes
the transpose operation. When the solid is considered to be incompressible, which is the case for
most biological tissues, det Fe = 1 for any deformation. Strain produces stress according to the
constitutive equation characterizing the mechanical properties of the solid. For an elastic medium
there exists a strain-energy density function W that depends on strain tensor e. Assuming that
the mechanical properties remain unchanged during the whole growth process and the material
is homogeneous, W is a function of e only and the Cauchy stress u defined per unit deformed
area in b(tl) is
aw T
t:r =Fe oe Fe +pi, (3)
where p is an unknown scalar function. In the absence of body forces the stress u satisfies the
equation of equilibrium formulated with respect to the current deformed grown configuration
234 A. Rachev
dm/dt
Figure 6. Schematic representation of growth rate-stress relation (Fung ( 1991)).
b(tl), \7 · u = 0 where \7 = fJ()jfJx(h). Boundary conditions are to be described for the

displacement defined as a difference between position vectors of the same particle in B 0 (t 0 ) and
b(h) and/or for the tractions on the bounding surface of b(tl).
Because the growth deformation gradient tensor F g is not in general known, additional in-
formation is necessary for the interrelations between F g and the tensors describing the strain
and stress field in the grown body. These functional relations are called the growth laws and phe-
nomenologically describe, at the continuum mechanics level, the effects of mechanical quantities
on the processes of mass supply or resorption.
Like any objective constitutive relation the growth laws have to be determined from simple
experiments or on the basis of comparison of model predictions for typical situations against
experimental observations. In general the growth deformation gradient tensor F g or its rate of
change may depend on the strain tensor e, the stress tensor u, the strain-energy density function
W or on the time rates of these quantities. It seems reasonable to consider strain as a proba-
ble candidate for a growth factor because changes in lengths and angles could be 'sensed' by the
cells. However, strains are defined with respect to the zero-stress configuration that a living tissue
never experiences under physiological conditions. Therefore postulating stress-dependent growth
laws was preferred. Such a growth law was proposed by Fung (1990) in the one-dimensional form
(4)
where dm/dt is the rate of mass growth, and C, a, b, c, k 1 , k2, and k 3 are constants to be deter-
mined experimentally. This relation, illustrated schematically in Figure 6 (Fung (1991)), shows
that stresses corresponding to the 'growth equilibrium states' do not initiate a change in the
mass.
The deviations of the stresses from 'equilibrium stresses' serve as driving stimuli for pro-
cesses resulting in increase or decrease of mass. Particular forms of growth law referring to the
remodeling of straight muscles, heart and arteries were proposed by Taber and Eggers ( 1996),
Taber (1998a) and Taber (1998b).
3.2 A Model for Aortic Growth Based on Fluid Shear and Fiber Stresses
To study growth and remodeling of aorta Taber (1998b) proposed a mathematical model using
the volumetric growth approach. It aims to check the hypothesis that changes in the aortic geom-
etry during development and in response to increased pressure are driven by the deviations of the
tensile stress from its baseline value and are modulated by the flow-induced shear stress.
Mathematical model. During the whole process of growth and remodeling the aorta was
considered to be a two-layered tube made of elastic orthotropic and incompressible materials,
representing the intima/media and adventitia. The vessel is inflated by an internal pressure P and
is extended longitudinally. It undergoes a volumetric growth resulting in an increase in length of
elements in the radial, circumferential and axial directions. The deformed grown state b(t!) at
the moment t = t 1 is related to the reference state B(t0 ) (Figure 7) by the relations
r = r(R, t), B = e, z = >.(t)Z, (5)

where>. is the axial stretch ratio 1 (R, e, Z) and (r, B, z) are the cylindrical coordinates of a point
in B( t 0 ) and b( t!), respectively. The total stretch ratios for each layer are
(6)
where Arg, Ae 9 , and Azg are the growth stretch ratios with respect to a cylindrical coordinates
(r, B, z) and>.;, >. 0, and>.; are the stretch ratio of elements lengths in the loaded state b(t!)
to lengths in the stress-free grown state (see Figure 6, configuration B'(tl)). Due to material
incompressibility >.; >. 0>.; = 1. The corresponding components of the Green strain are ei =
e,
1/2 (>.i 2 - 1), i = (r, z).
Constitutive equations for the wall material relate the stresses in b( tl) to the Green strain via
CJi = >.*2 aw + P (7)

" ae;:
where W(e;, eiJ, e;) is the strain-energy density function andp(r, t) is an unknown scalar func-
tion to be determined from the equilibrium equations and boundary conditions.
The radial equilibrium equation is
ocrr err -ere _ 0
ur
!l + r - ' (8)
and the boundary conditions at the inner surface (r = ri) and outer surface (r = r 0 ) are
(9)
Continuity of cr r on the contact surface between two layers was imposed as well. The vessel
is considered to have closed ends and the overall equilibrium in the axial direction yields
2lro CJzrdr = rf P. (10)
1 Strain and stress analysis of a thick-walled tube is given in Appendix A. Because of coupling of the
deformation and growth processes the basic field equations are given here for completeness.
236 A. Rachev
(a) (b) (c)
Figure7. Schematic representation of(a) reference state B(t0 ); (b) deformed grown state b(t 1 ); and (c)
opened-up configuration.
Based on experimental observations that sustained changes in flow cause a change in radius,
while persistent changes in pressure mainly affect wall thickness, Taber postulated the following
stress-based growth laws
(a-9-
8Agr _ __!__
8t - Tr
a-9o)
(a-9o)m '
(11)
8>.9e (a-9- a-90) (T- To)

1
Tt = Te (a-eo)m + Tr (To)m
1 (12)
8Agz = Q
at ' (13)
where T is the flow-induced shear stress on the endothelium; Ti are time constants; a is a con-
stant characterizing the decay of the signal with distance from the inner surface; a-eo and To are
the circumferential wall stress and flow-induced shear stress; the subscript m denotes values at
maturity. It was assumed that the growth-equilibrium stresses depend linearly on blood pressure
during development, and then remain constant in the matured organism. At t = 0 the growth
stretch ratios >.9 r, >.9 e, and Agz were set to be equal to one, i.e. no growth has occurred.
Results and Discussion. Prescribed time courses of mean pressure and blood flow rate are
illustrated in Figure 8 for the period of development and maturity of a rat aorta and for a period
of sustained increase in pressure and flow by 50%.
After specifying the strain-energy function and all model parameters the governing equations
were solved numerically. The results obtained showed realistic time courses of geometrical di-
mensions of the rat aorta during development and under hypertensive conditions (Figure (9)).
The model predictions are in agreement with the experimental data of Liu and Fung (1989) and
support the hypothesis that mechanical forces have significant effects on the geometry of the
aorta. The model suggests that the changes in aortic geometry might be considered to be driven
by the tensile wall stress and to be modulated by the flow-induced shear stress according to the
growth law (eq. (13)).
(a) (b)
30 3000 Normal
";' -------- Perturbed
""'
u
~ 20 Ill
2000 --------
1
/
<Jl
I
~
"'"' 10
....
(!)
p., ~ 1000
0
G:
0 0
0 100 200 300 100 200 300
0
Time (days) Time (days)
Figure 8. Prescribed time variation in (a) blood pressure and (b) blood flow. First heart beat: t = 0; birth:
R = R/ R 0 ; maturity: t ~ 100 days; (Taber (1998b)).
200 (a) 10 (b)
bil 150
J8
(!)
~ ~ 6
"'
eo.:
(!)
100 ~ 4
<t: "'
~
50 2
0 0
0 100 200 300 0 100 200 300
Time (days) Time (days)
Figure 9. Computed time course of (a) opening angle and (b) geometrical dimensions. R1 and R2 are the
inner and outer radius in the unloaded state; H is the wall thickness. Pressure increases by 50%
at t = 200 days (Taber (1998)).
3.3 Global Growth Approach

The global growth approach was proposed for studying geometrical changes in arteries caused by
changes in arterial pressure (Rachev eta!. (1996) and Rachev eta!. (1998)). The basic hypothesis
is that remodeling results solely in a change in the geometrical dimensions of the zero-stress
configuration. The rates of change of geometrical dimensions and the parameters that describe
the strain and/or stress state of the arterial wall are related by appropriate evolution equations.
Referring to the case of the volumetric growth described in the previous section, this approach
assumes that all changes of differential elements are mutually compatible and no deformation
is required to assemble the grown elements into a continuum body. Configurations of a growing
artery are shown schematically in Figure 10.
The mapping of the original zero-stress configuration B(t 0 ) into a grown zero-stress config-
uration B' (t 1 ) represents a homotopic transformation. The global growth approach is a particular
case of the more general theory of volumetric growth and therefore has limited applicability.
However, there are at least two cases when the use of this approach is justified. First, when an
artery is considered as a thick-walled tube, but it is assumed that a single radial cut releases all, or
238 A. Rachev
Deformation and Remodeling
B(to) b(to)
Observed
) .7 0
growth
====>
cr>'O
\' Growth
Geometry Evolution
~
cr=O
Figure 10. Schematic representation of configurations resulting from deformation and global growth. B(to)
is the original zero-stress state; b(to) is the instantaneous deformed state; B'(t1) is the grown
zero-stress state; b( h) is the grown deformed state.
at least most, of the residual stress and the opened-up configuration is stress-free. Second, when
an artery is considered to be a thin-walled elastic membrane. As shown in Appendix B in this
case the state of no load is stress-free.
To illustrate the global growth approach models of arterial adaptation to sustained changes in
pressure (Rachev et al. (1998)) and flow (Rachev (2000)) are considered.
3.4 A Model for Dynamics of Adaptation of Arteries to Induced Hypertension
In order to test certain hypotheses concerning the driving stimuli and the interrelation between
shear-dependent, wall stress-dependent and compliance-dependent processes during arterial adap-
tation a phenomenological mathematical model has been proposed (Rachev et at. (1998)).
Mathematical model. The artery is considered to be a thick-walled tube made of nonlinear

elastic orthotropic and incompressible material. At the zero-stress state, the geometry of the
vessel cross-section is considered to be a circular sector with the following dimensions: inner
arc length, L;, outer arc length, L 0 , and thickness, H, (Figure lla). The inner and outer radii of
curvature and the opening angle are calculated using the formulas
6H cJ> _ _ Lo- Li
2H ' (14)
Ro=6-1' - 7r
where 6 = L 0 / Li.
The elastic properties of the vascular tissue are described by a strain-energy function pro-
posed by Chuong and Fung (1986) in the form
c exp[b1ee2
W = 2 + b2ez2 + b3er2 + 2b4eeez + 2b5ezer + 2b6ereo l, (15)
(a) (b)
\ 4
· ~j ~
--r
;.<.
<D ••• •• \
-------~--
.: :. . . ·· ~ I
;---
!.__ _ _____,~ ,
Figure 11. Schematic representation of(a) reference state B(to), and (b) deformed grown state b(t1).
where ei, i = (r, (}, z) are Green strains and c and bj are material constants. Under physiological
loads the vessel undergoes a finite inflation and finite axial extension. The stress and strain state
in the arterial wall is described in Appendix A.
Induced arterial hypertension is modeled by a step increase in blood pressure from pN to
pH. The magnitude of the blood flow was kept constant. To monitor remodeling of the zero-
stress state configuration the following growth parameters were defined:
L{j (t) HH(t)

/3( ) = "t(t) = HN , (16)
t LN '
0
where the superscript N and H refer to the normotensive and hypertensive artery, respectively.
Remodeling rate equations. The evolution of the geometrical parameters of arterial cross-
section was described by the following remodeling rate equations for the growth parameters
where u(l (t) , u[! (t), uf/ are the current circumferential stress at the inner surface, the stress at
the outer surface, and the average circumferential of the hypertensive artery; u{", u{;, u{/ are the
corresponding stresses in the normotensive artery; 7 1 , and 7 2 are characteristic time constants.
Equation ( 17)1 yields the rate of change of the inner arc length of the arterial cross-section at
the zero-stress state of the hypertensive artery. The term in parentheses represents the normalized
deviation of the circumferential stress at the inner arterial surface and at the current hypertensive
state from its value under normotensive conditions. Equation ( l7)z is analogous to eq. ( 17 1 ) but
refers to the outer arc length. Equation (17)3 describes the evolution of the arterial thickness,
which was accepted as being driven by the deviation of the average circumferential stress at
the hypertensive conditions with respect to the normotensive value. Time constants 7 1 to 7 2 are
thought to be different because they relate to different types of remodeling.
A step increase in the blood pressure causes a step increase in the inner radius of the artery in
the deformed state. Thus, although the flow is accepted as constant during the wall remodeling,
240 A. Rachev
in accordance with eq. (2) the shear stress at the inner arterial surface also undergoes a step
increase. To take into account the experimental fact that the arterial remodeling process restores
the wall shear stress to near normal level (Langille (1993) and Zarins et al. (1987)), the following
equation is postulated for the evolution of the inner deformed radius of the hypertensive artery
(18)
where T 3 is a time constant related to flow-induced remodeling of the inner arterial diameter. The
term in the brackets on the right hand side of eq. ( 18) is the normalized deviation of the flow-
induced shear stress from its value under normotensive conditions. To keep the calculated values
of a, j3 and"( compatible with values which give the solution of the evolution eq. (18), a scaling
procedure was introduced based on the assumption that at any moment of the geometrical adap-
tation of the hypertensive artery its deformed inner radius follows the compensatory adjustment
driven by the flow-induced shear stress (see Rachev et al. (1998)).
It follows from eqs. (17) and (18) that remodeling of the zero-stress configuration of the
hypertensive artery lasts until a steady state is reached. Since the circumferential stress varies
smoothly throughout the arterial wall, equal values of the average stress and the stresses at the
inner and outer surface of the normotensive and hypertensive arteries imply an equivalent distri-
bution of the circumferential stress with respect to normalized wall thickness.
The pulsatile nature of the cardiac output imposes a cyclic pressure on the arterial wall and
the artery undergoes a small pulsatile circumferential deformation relative to the deformed state
caused by the mean pressure and in situ axial extension. The cyclic strain in response to a pulse
pressure LJ.P is determined by the arterial compliance defined as C = LJ.ri/riiJ.P , where n is
the inner radius at mean pressure and in situ length, and LJ.ri is the change in radius. The compli-
ance characterizes the overall mechanical response of an artery to small physiological variations
of the pressure around its mean value. Due to non-linear dependence of radius on mean pressure,
a step change in pressure affects the compliance of the artery. It was postulated in this study
that remodeling of the mechanical properties serves to restore the arterial compliance as exists
under normotensive conditions and thus restores the normal function of the artery as an elastic
reservOir.
Assuming that the analytical form of the strain-energy function remains unchanged, a change
in the mechanical properties affects the values of the material constants. It was shown that the
material constants c, b1 , and b2 provide the main contributions to the nonlinear stress-strain rela-
tionships in the circumferential and axial direction, respectively (Fung and Liu ( 1991)). Therefore
the variation of the mechanical properties induced by the sustained hypertension was studied in
terms of alteration ofthe material constants c, b1 , and b2 , considering the other material constants
to be unchanged.
Thus, the following evolution equation describing the mechanical adaptation was introduced
d~(t)
(19)
dt
where ~ is a material growth parameters defined by the relation
(20)
1.0 130
0.9 Oil
a (\)
~ 120
5 0.8
(\)
"5lJ
"'"'
"'=
(\)
,.,..= OJ)
:.aE-<
0.7
·a=
(.)
110
(\)
0.
0.6 0
0.5 100
0 10 20 30 40 0 10 20 30 40
Dimensionless time Dimensionless tirre
Figure 12. Predicted time course of (a) wall thickness and (b) opening angle.
and T 4 is a time constant.

Allowing for the experimental observations that the geometrical adaptation to increased pres-
sure precedes the mechanical adaptation, it was hypothesized that at any moment of arterial
remodeling the changes in the mechanical properties of vascular tissue maintain the achieved
stress distribution in the arterial wall caused by the geometrical remodeling. Therefore the cur-
rent values of the material constants b1 and b2 for the hypertensive artery were calculated from
the equations
where D { ... } /Dt denotes the total derivative;

mean axial stress, respectively.
a: a:
and are the mean circumferential and
Based on the definition of the growth parameters the initial conditions at time t = 0, when
the pressure increases in a step-wise manner from pN to pH, are a = (3 = 1 = ~ = 1.
Equations ( 17) and ( 19) are nonlinear because the wall stresses depend in a complex manner
on the geometrical parameters of the current zero-stress state and on current mechanical proper-
ties. The equations are coupled to the field equations of the artery subjected to the hypertensive
pressure. A numerical study was performed using the experimental data for a rabbit thoracic aorta
given by Chuong and Fung (1986). Hypertension was simulated by a step increase in pressure
from 100 mmHg to 160 mmHg. Time constants were chosen to scale the course of the geomet-
rical and mechanical parameters in keeping with available experimental data of Fung and Liu
(1989) and Matsumoto and Hayashi (1994).
The numerical results obtained show a monotonic increase in the wall thickness of the hy-
pertensive artery (Figure 12a), a fast restoration of the deformed inner radius (not shown) and
non-monotonic time course of the opening angle (Figure 12b). These results are in qualitative
agreement with experimental observation ofFung and Liu (1989) and Liu and Fung (1989).
Remodeling of mechanical properties was considered from two aspects. First, the changes
that occur in the mechanical properties of the wall material, and second the changes in the me-
chanical response of the artery to applied loads. Figures 13a and 13b illustrate the time course of
the material constants C and b1 in the strain-energy function given by eq. (15) and thus reflect
the variations that appear in the stress-strain relations in the vascular material. Because the prod-
242 A. Rachev
80 (a) (b)
1.2
(? •N
0.. f
c 60 '0 1.0
}.> §
0.8
=
tl 40 f
"'
=
0
(.) =
tl 0.6
'"
·c
~
20 =
0
(.)
0.4
::E '"
·c
~ 0.2
0 ::E
0 10 20 30 40 0 10 20 30 40
Dimensionless time Dimensionless time
Figure 13. Predicted time course of(a) material constant C and (b) material constant b1 and b2 .
uct Cb 1 is proportional to the slope of the circumferential stress-strain relation in the zero-stress
state ( -\e = Az = 1), the model predicts that the arterial tissue of hypertensive artery is stiffer in
the zero-stress state than its normotensive counterpart. Pressure-radius relationships (Figure 14)
represent the mechanical response of the blood vessel incorporating both the inherent elastic
properties of the vascular material and the geometrical dimensions in the zero-stress state. At
any pressure level the deformed inner radius of the hypertensive aorta is smaller than the radius
of the normotensive vessel, as was found experimentally by Vaishnav et al. (1990).
The proposed model describes phenomenologically the process of arterial wall remodeling
disregarding details about differentiation or hypertrophy of cells, production of extracellular ma-
trix, and changes in the composition and arrangement of basic structural components of the
arterial wall. Agreement of the model predictions with most of the available experimental data
supports the hypothesis that geometrical adaptation might be considered as driven by the devi-
ation of the wall stress from its baseline values while the mismatch of the arterial compliance
plays the role of a driving stimulus for adaptation of the mechanical properties.
The global growth approach has been used to account for the changes observed experimen-
tally by Fridez et al. (200 1b) in the muscular tone in response to induced hypertension. The study
addresses the hypothesis that the synthetic and proliferate activity of smooth muscle cells, lead-
ing to changes in the dimensions of an artery exposed to sustained hypertension, is associated
with changes in the contractile state of the smooth muscle cells and changes in the circumferen-
tial wall stress. The model has the merit of being the first one based on a single comprehensive
data set. It is robust to changes in parameter values within the physiological range, and yields
quantitative information on adaptation processes under hypertensive conditions. The results ob-
tained contribute to better understanding of the normal arterial function and reveal interrelations
between different biomechanical events that are associated with arterial remodeling.
An approach in between the volumetric growth (Taber and Eggers (1996) and Taber (1998b))
and the global growth (Rachev et al. (1996) and Rachev et al. (1998)) was used to study the
effects of sustained hypertension on arteries when geometrical remodeling is complete (Rachev
( 1997) ). The arterial wall was considered as a two-layered tube. Each layer remodels to maintain
the stress distribution, as it exists in the normotensive artery, by changing the geometrical dim en-
1.2 1.6 2.0 2.4 2.8

Internal radius ri (mm)
Figure 14. Pressure-radius relationship for normotensive and hypertensive artery.
sions of the sectorial zero-stress configuration. The zero-stress configurations of the layers of the
adapted hypertensive artery are not geometrically compatible and the opened-up configuration,
which appears after a radial cut in an unloaded ring segment, is not stress free but contains resid-
ual stresses.
Despite encourageing predictive results the models based on the global growth approach need
further improvement and experimental verification. Supporting motivation for the proposed evo-
lution growth equations and the hypotheses introduced for the compatibility of the stress-induced
geometrical adaptation, compensatory adjustment of the inner radius and changes in mechanical
properties of the arterial tissue are yet to be found. More experimental data for the time course
of geometrical and mechanical remodeling are needed for better identification of model parame-
ters. A generalization of the model might consider the arterial wall as a multi-layer structure with
non-homogeneous mechanical properties and different rates of remodeling of the basic structural
layers, as was suggested by Fung et al. (1993).
3.5 A Model of Arterial Adaptation to Changes in Blood Flow
Experimental investigations have shown that the arterial response to changes in blood flow in-
volves processes in which the smooth muscle cells play a key role. Therefore, relevant modeling
should include both the short- and long-term contribution of the vascular smooth muscle. A math-
ematical model for both the acute vasomotor response and the long-term geometrical remodeling
of arteries induced by sustained changes in blood flow was proposed by Rachev (2000). The
model aimed to give a probable interpretation of some experimental results available in the lit-
erature and to suggest new types of experiments. The study addressed the hypothesis that the
synthetic and proliferative activity of smooth muscle cells, leading to a change in arterial di-
mensions, is shear stress dependent and is associated with changes in the contractile state of the
244 A. Rachev
smooth muscle and changes in the circumferential wall stress.
Mathematical model. The artery was considered to be a thin membrane made of nonlinear
elastic incompressible and orthotropic material. Based on experimental findings that in some
cases an artery remodels without altering significantly its structure and composition (Langille
and O'Donnell (1986) and Fath et al. (1998)) mechanical properties were considered unchanged
during the whole remodeling process. Under physiological loads the artery undergoes a finite
elastic deformation (see Appendix B). To describe the stress state the approach used by Rachev
and Hayashi (1999) was adopted. On the basis of Hill's functional model (see Hill (1939)) the
total circumferential wall stress per unit deformed area, ITT, was represented as
(22)
where ITp is the passive stress that is borne by the wall material when the vascular smooth muscle
cells are fully relaxed and >.o is the circumferential stretch ratio. The second term in eq. (22)
represents the active stresses developed by SMCs when stimulated.
The active stress depends on several factors. It is well recognized that the magnitude of the
active tension (the stress resultant of the active stress) developed at a constant arterial diameter
(at so-called isomeric constriction) and constant stimulus depends on the actual diameter. There
exists an optimum deformed diameter at which the active tension developed by the SMCs has
maximum value, while below and above certain values the active tension is zero. The function
J(>. 0 ) in eq. (22) is a normalized function, which accounts for this diameter-tension relationship
(Dobrin (1973)). On the other hand, at fixed length and variable stimulus the active tension de-
pends on the intensity of stimulation following so-called active tension-dose relationship. This
relationship reflects the level of the contractile activity ofSMCs, which in tum depends on the in-
tercellular concentration of free calcium ions. It has been established that at the homeostatic state
the relationship between the developed active tension and free calcium concentration is close to
sigmoidal with a pronounced linear portion over the physiological range of stresses (Rembold
and Murphy (1990)). Finally, the active tension depends on the amount and orientation of the
smooth muscle cells in the arterial wall. The parameter S in eq. (22) represents the maximum
active stress per unit undeformed area (nominal stress) for given intensity of muscular stimula-
tion at a stretch ratio giving f(>.omax) = 1. The factor >.o in eq. (22) accounts for the fact that
the actual active stress (Cauchy stress) is defined per unit deformed area.
Accounting for the active circumferential stress the equation of overall equilibrium of the
vessel in the radial direction is (see Appendix B, eq. (B6))
(23)
where Pis the applied arterial pressure, Az is the axial stretch ratio and Rand H are the mid-wall
radius and wall thickness at the state of no load and no muscular tone (S = 0).
Evolution equation for the contractile state of vascular smooth muscle. The artery was
first considered under conditions at which the pressure, flow and muscular tone parameter have
their baseline values P0 , Q0 , and 50 , respectively. The mid-wall radius and wall thickness in the
unloaded state are denoted by Ro and H 0 . Hereafter, the subscript 'zero' will be used to refer to
the artery under normal flow conditions. The mean flow rate was changed in a step-wise manner
from Q0 to Q* and kept constant, while the pressure remained constant at its baseline value.
At the moment of the flow jump, the vessel maintains its inner radius at the control value no.
According to eq. (2) the change in the flow rate causes an instantaneous jump in the shear stress
at the arterial lumen from To to T1 = To(Q* /Q 0). Over time the magnitude of the flow-induced
shear stress varies as a result of changes in the inner radius caused by the vasomotor response
and geometrical remodeling.
Exploiting the experimental observations that for mature vessels the amount of major struc-
tural components of the wall material remains unchanged (Langille and O'Donnell (1986) and
Fath et al. (1998)), it was assumed that the ratio of the area occupied by the smooth muscle cells
to the total area remains constant during the whole process of remodeling. Therefore, the mus-
cle parameter S depends only on the calcium concentration within the cells. The equation that
describes the dynamics of S was deduced from the processes that control the calcium concentra-
tion.
The basic mechanisms which lead to increase of ca++ concentration are: i) incoming flux
of calcium ions due to concentration drop between the cell and its environment, and ii) release
ofCa++ from intercellular stores such as the sarcoplasmatic reticulum, where the ions are in the
weakly bound form. The type and intensity of the applied stimulation affect both processes. On
the other hand, there exist simultaneous processes, which cause a decrease in Ca ++ concentra-
tion: active transport outside of the cell by so-called calcium pumps, sodium-calcium exchange
mechanisms, and resorption ofCa++ in the sarcoplasmic reticulum. Following the phenomeno-
logical approach proposed by Rachev et al. (1980) and its modification given by Achakri et al.
(1994), the rate of change of calcium concentration in the smooth muscle cells is described by
the balance equation
de 1
- = t.p-- c. (24)
dt T
The non-negative function t.p represents the processes that lead to an increase in calcium
concentration. The value of t.p depends on the applied stimulus reflecting the coupling between
stimulus and transmembrane flux through receptor-operated channels or release of calcium from
intracellular stores. The second term on the right hand side of eq. (21) accounts for the decrease
of calcium within cells, which is assumed to be proportional to the actual Ca ++ concentration,
T being a positive constant that gives the decay rate of calcium concentration.
It was assumed that the deviation of t.p from its baseline value is proportional to the devia-
tion of the shear stress, i.e. t.p- t.po = -(3(T- To), where f3 is a coefficient of proportionality.
This relation phenomenologically reflects the effects of altered shear stress at the endothelium
on SMCs membrane permeability and internal calcium sources. It follows from eq. (22) that an
increase in shear stress over its baseline value causes a decrease in intercellular Ca ++ concen-
tration resulting in relaxation of the vascular smooth muscle. A decrease in shear stress under
its baseline value has an opposite effect on Ca ++ concentration and leads to contraction of the
smooth muscle cells.
Using the linear dependence of the active tension on Ca++ concentration
(Rembold and Murphy (1990)) and introducing a dimensionless variable, the equation for the
evolution of the muscle parameter S describing the ionic state of smooth muscle cells takes the
246 A Rachev
form
ciS= _1 ( 1 - ~) + _1 (S _ 1), (25)
dt Tst To Ts2
where S = S ISo, while Ts2 and Ts2 are time constants.
Remodeling rate equations for the case of increased flow. Following the global growth ap-
proach dynamics of wall remodeling was described by evolution equations for the mid-wall ra-
dius and wall thickness in the unloaded state. Generalizing experimental observations for growth
and remodeling of tissues, Rodbard (1970) has proposed that tension applied to relaxed muscle
causes its longitudinal growth. Allowing for this hypothesis the compensatory enlargement in
response to increased flow was assumed to be driven by the increased passive circumferential
stress while the muscle is relaxed. On the other hand the increase in stress elicits transversal
remodeling, which provides sufficient wall thickness to restore the normal value of the total cir-
cumferential stress. Introducing the dimensionless variables the corresponding remodeling rate
equations are
(26)
where R = Rl R 0 , fi = HI H 0 , TR and THl are time constants and uT is the total circumferen-
tial stress.
Remodeling rate equations for the case of reduced flow. Geometrical remodeling in re-
sponse to reduced flow develops mainly as intimal thickening. It was postulated that cellular and
intracellular processes, which result in thickening of the arterial wall are associated with the in-
creased contractile activity of the smooth muscle cells. The following evolution equation for the
dimensionless wall thickness was proposed
dii 1 -
- = -(S-1), (27)
dt TH2
where TH2 is a time constant. For deformations in the physiological range the active tension de-
veloped by the smooth muscle cells increases when S increases (see Rachev (2000)). Therefore,
eq. (27) is consistent with the hypothesis of Rodbard (1970) suggesting that increasedcontract-
ing tension causes the muscle to thicken.
The evolution equation for the change in mid-wall radius is based on the experimental obser-
vations that mature vessels exhibit practically no change in vessel mass as a result of structural
remodeling in response to reduced flow (Langille et al. (1989) and Langille (1993)). Moreover,
adaptation is achieved without altering the vessel length (Langille (1993) ). Because the wall
material is considered incompressible, a constancy of mass implies a constant arterial cross-
sectional area at the state of no load. Considering this finding, it was assumed that the area of the
undeformed cross-section remains unchanged during the entire remodeling process and hence
fill= 1.
The evolution eq. (25) and the remodeling rate eqs. (26) and (27) represent a system of non-
linear autonomous equations. They are coupled to the equations that describe the deformation
and the equilibrium of the artery, and the equation for the flow-induced wall shear stress. The
1.1 (a) 1.4 (b)

3
1.0
4 1.3
gj
5
] I
1
0.9 1.2
2 21
/
1
0.8 1.1
4
0.7 1.0
:i:
0.6 0.9
0 2 4 6 8 10 10000 20000 30000 40000 50000 0 2 4 6 8 10 I 0000 20000 30000 40000 50000
Time (minutes) Time (minutes)
1.25 (c) 2.0 (d)
gj
1.8
]
~
4
1.00 1.6
!i
]
.g 2
1.4
I
0.75 1.2
3 1.0
0.50 0.8
0 2 4 6 8 10 I 0000 20000 30000 40000 50000 0 2 4 6 8 10 10000 20000 30000 40000 50000
Time (minutes)
Time (minutes)
Figure15. Time course of(a) normalized initial mid-wall radius R = RIRo; (b) normalized initial wall
thickness fi = HI Ho; (c) normalized deformed inner radius i'i = T'i Ir;o; (d) normalized maxi-
mal active stress S = S ISo after increasing blood flow by factors of two (line 1), three (line 2)
and five (line 3). Line 4 refers to the case of a twofold increase in flow and a tenfold increase in
the time constant Tsl·
initial conditions that correspond to the time t = t 0 of the onset of the step change in the flow
are
S(to) = 1, R(to) = 1, H(to) = 1. (28)
Results and Discussion. The governing equations of the process of arterial adaptation were
solved numerically for model parameters identified from experimental data for a canine carotid
artery (Cox (1978)). Figure 15 illustrates the time course of the dimensionless initial mid-wall
radius, thickness, deformed inner radius, and the smooth muscle parameter representing the max-
imal active Lagrangian stress. The non-monotonic pattern of variation of S (Figure 15d) and
stretch ratio >.o (not shown) as well as the constancy of the undeformed arterial thickness and
mid-wall radius during the early period of flow alteration (Figures 15a and 15b) show that the ves-
sel response is initially due to arterial dilatation. As geometrical remodeling proceeds, however,
initial dimensions increase slowly and monotonically to their adapted values. Simultaneously,
the muscle cells restore their normal ionic state ( S = 1) and the circumferential stretch ratio
returns to its baseline level. Hence, the vessel restores both the passive and active stress, as they
exist under normal flow conditions.
The time variations of the geometrical and mechanical parameters following a reduction in
flow are shown in Figure 16. The results again show a monotonic change in the wall dimensions
248 A. Rachev
1.1 (a) 1.4 (b)
3
.a 1.0
4 1.3
-o
~ 2
1
~ 0.9 1.2
2
~
-o 0.8 3 1.1
1:l
~ 4
§
0 0.7 1.0
z
0.6 0.9
0 2 4 6 8 10 I 0000 20000 30000 40000 50000 0 2 4 6 8 10 I 0000 20000 30000 40000 50000
1.25 (c) 2.0 (d)
~
:: 1.8
;:; 4
~
~
.5 1.00 "'
>
·a 1.6
-o g
§"'
<8 ~
-o
1.4
.1J
~
] 0.75
"'§
1.2
~0
3 0
z 1.0
z
0.50 0.8
0 2 4 6 8 10 I 0000 20000 30000 40000 50000 0 2 4 6 8 10 I 0000 20000 30000 40000 50000
Figure16. Time course of(a) normalized initial mid-wall radius R = R/Ro; (b) normalized initial wall
thickness fi = H / H o; (c) normalized deformed inner radius Ti = ri/ Tio; (d) normalized maxi-
mal active stress S = S /So after reducing blood flow by factors of two (curve I), three (curve 2)
and five (curve 3). Curve 4 refers to the case of a 50% reduction in flow and a tenfold increase in
the time constant Ts 1 .
and deformed inner radius. The primary response to reduced flow is vessel constriction due to
increased vasomotor tone. Geometrical remodeling again restores the control level of the smooth
muscle cells ionic state. In contrast to the case of arterial enlargement, however, the circum-
ferential stretch ratio reaches an asymptotic value, which is smaller than the control value (not
shown). Therefore, the passive and active stresses in the adapted vessel become smaller than the
corresponding control values.
Shear deformation of the endothelial cells appears to be the first in the chain of events that re-
sult in smooth muscle relaxation or contraction. The ability to sense the deviation of shear stress
from its normal level might be impaired due to endothelial dysfunction or decreased deformabil-
ity of the endothelial cells. Theoretical predictions of the model for the case of large values of the
time constant Ts 1 , simulating very slow or even missing sensitivity to flow-induced shear stress,
showed that the ionic state in the smooth muscle cells remains practically unchanged (curves
4 in Figures 15 and 16). Therefore, neither fast vasomotor adjustment of the arterial radius nor
geometrical remodeling becomes apparent. The conclusion that wall adaptation is endothelium
dependent is in agreement with experimental observations for missing flow-induced response in
arteries denuded of endothelium.
So far, a deviation of flow-induced shear stress from its baseline value has been recognized
as the single mechanical parameter that serves as a mediator between changes in blood flow
and geometrical remodeling. An abrupt change in shear stress, however, could rapidly diminish
almost completely as a result of a vasomotor adjustment of the arterial lumen to altered flow con-
ditions. It seems reasonable to seek interrelations between rate of change of arterial dimensions
and other mechanical values that originate from flow changes and persistently remain altered
while the artery remodels. The basic hypothesis in this study is that geometrical remodeling in
response to altered blood flow is associated with sustained changes in the contractile state of the
smooth muscle cells, which is modulated by the flow-induced shear stress sensed by the endothe-
lial cells. The altered contractile activity affects the stress and strains experienced by the smooth
muscle cells and might thereby enhance their secretory function as well as the pattern and the
rate of cell growth and division.
The model describes both the acute vasomotor response and the chronic geometrical remodel-
ing that follow a change in blood flow. The model predictions that remodeling is a stable process
and results in restoration of the baseline values of the wall shear stress under normal condi-
tions are in agreement with experimental observations available in the literature (Brownlee and
Langille (1991) and Langille (1995)). The results obtained also showed that the ionic state of the
vascular smooth muscle cells reverts to the normal homeostatic state and the muscle restores its
contractile activity to the control level, which is in keeping with the recent findings of Hayashi
(2000).
In summary, this study proposed a relatively simple mathematical model in terms of contin-
uum mechanics. The model does not account implicitly for the series of interrelated electrical,
chemical, mechanical and biological processes involved in smooth muscle contraction and re-
laxation, regulation of vascular cell mitosis and apoptosis rates, cell migration, control of matrix
synthesis and degradation, and regulation of matrix reorganization. The study focused on three
main events that occur after a change in blood flow: i) a change in the ionic state of smooth mus-
cle cells caused by the altered shear stress sensed by the endothelium; ii) a change in the stress
state in the arterial wall due to altered muscular tone; and iii) a change in the geometrical dimen-
sions of the arterial cross-section due to remodeling. The model suggested plausible hypotheses
for arterial wall remodeling and predicts the main features of arterial response to changes in
blood flow.
The thin-wall membrane model used in this study disregarded the differential growth across
the wall and the contribution of residual strains. Further generalizations of the model should con-
sider arteries as thick-walled tubes and the effects of residual stress should be taken into account.
It is also necessary to verify experimentally the assumption that the cross-sectional area of the
arterial wall remains constant throughout the entire remodeling process induced by a chronic
reduction in blood flow, or whether it returns to its original value at the end of the process.
Modified equations for evolution of muscular tone and geometrical dimensions and mechanical
properties are needed for modeling flow-induced remodeling of arteries during development and
maturation. In general, there is a need for additional data on the time variation of geometrical
and mechanical parameters during remodeling.
Finally, it is worth noting that in all the models that were described in the section Mathemat-
ical Models of Remodeling and were based on the volumetric growth or on the global growth,
the processes leading to a change in arterial size and mass were considered only in the context of
their kinematics by postulating growth laws or remodeling rate equations. More adequate mod-
250 A. Rachev
eling must consider the coupling between the deformation of the arterial wall and the processes
of mass transport. Such an approach requires consideration of the arterial tissue as a multiphase
continuum. A general model of growing biological tissue consisting of solid and fluid phases
and, in particular, application to growth of plants was proposed by Stein (2000).
4 Specific Biomechanical Problems Associated With Arterial Remodeling
Vascular diseases and, in particular, the coronary artery disease are a leading cause of death in
many countries. Substitution or bypassing of a stenosed part of an artery by a vascular graft,
or performing a balloon angioplasty followed by a stent deployment are common procedures
in vascular surgery. Unfortunately, in many cases the treatment is followed by an undesirable
occlusive response called restenosis. Restenosis is a result of neointima formation, termed intimal
hyperplasia, and also partly a result of proliferation of the vascular smooth muscle cells in the
media. Because graft or stent deployment affect the mechanical environment of the treated artery,
it is reasonable to consider restenosis as an outcome of a remodeling response. Understanding
the origin of this process and the factors on which it might depend could promote better design
and selection of arterial grafts and stents and may suggest improvements in surgical techniques.
Some mechanically-based studies in that respect are discussed in this section.
4.1 A New Principle to Control the Flow-Induced Remodeling in an Arterial Segment
Restenosis due to intimal hyperplasia occurs at a high rate after balloon angioplasty followed by
a stent deployment and is the major negative outcome that compromises the therapeutic inter-
vention. There exists a great deal of evidence that intimal hyperplasia is initiated and modulated
by the reduction of wall shear stress. It seems reasonable to accept that arterial restenosis is a
compensatory response tending to restore the baseline value of the wall shear stress (Glagov et
al. (1993)). A strategy for preventing restenosis could be an appropriate increase in the local wall
shear stress within a portion of an artery where a restenosis may occur. Because it is not possible
to keep locally an increased blood flow, an alternative approach to affect the shear stress is to alter
the velocity profile inside the arterial region. An original idea in that respect has been proposed
by Stergiopulos (2000). The keystone is to insert a small cylindrical body in the streaming line
of the flow that disturbs the velocity but does not affect significantly the local flow and the hemo-
dynamic resistance. This principle is illustrated in Figure 17. An insert with diameter one-third
of the arterial lumen causes a two-fold increase in the wall shear accompanied by a negligible
augmentation of the total resistance. Recent animal experiments with an ARES stent designed
and manufactured on this principle confirmed its suitability and usefulness in preventing arterial
restenosis.
4.2 Shear-Controlled Competition Between Collateral Vessels
It has been observed that the number of vessels that supply the same tissue and run in parallel,
the so-called collateral vessels, tends to decrease. This tendency clearly manifests itself during
embryonic development when the number of large arteries and veins declines. For example, the
large number of arteries that supply blood to the kidney gradually decreases to one or two vessels
to perform this function (Kamiya and Togawa (1980)). From a physiological point of view the
•• :•••••••••••••••••m••m••• [ ) •••••••••••••••••••••• : •••
Figure 17. Schematic representation of velocity profiles in the absence and in the presence of a cylindrical
insert (Stergiopulos (2000)).
reduction of the number of vessels is of importance because it diminishes the amount of tissue
required to form the vessels and the space for their disposition. Large vessels do not contribute
significantly to the energy losses in the vascular system and arterial pressure, but in any case
fewer arteries leads to a reduction of the cardiac work.
A competition of two vessels running in parallel that lead to occlusion (closure) of one of
them has been observed after bypass implantation. When a stenosed (narrowed) artery is by-
passed by an arterial graft it has been found that the stenosis keeps on progressing and leads to a
total occlusion (Albridge and Trimble ( 1971)).
Keenan and Rodbard (1973) proposed an explanation of these observations based on the
mechanisms adopted for arterial remodeling in response to changes in blood flow. It was hy-
pothesized that arteries enlarge or diminish their deformed inner radius to keep the flow-induced
shear stress at the inner surface at the baseline value. The major result is that the configuration of
two vessels that bifurcate from one vessel and reunite again is not in a stable homeostatic state.
The vessels are shown schematically in Figure 18. The total flow Q is the sum of the individ-
ual flows Qa and Qb
(29)
where the subscripts a and b refer to the vessels in Figure 18. Assuming Poiseuille flow and
constant pressure drop
(30)
4Qf.1
Tm = - 3 - , m =(a, b), (31)
7rrm
where rm is the deformed inner radius.
Allowing for the fact that increased shear stress over the baseline value induces arterial en-
largement while decreased stress makes the vessel to reduce its lumen, the following evolution
equations were proposed
drm 1( )
dt = T Tm- Tc' m =(a, b), (32)
where Tc is the baseline value of the wall shear stress and T m is its current value, T being a time
constant.
Substituting eqs. (31) into eqs. (32) and making use of eqs. (30) yields
(33)
252 A. Rachev
Figure 18. Schematic of competing two-vessel system (Keenan and Rodbard (1973)).
where k = 4Q Jl,j 7f.

Equations (33) represent a system of two autonomous equations, called the dynamic system.
It has a stationary solution, obtained by setting the left-hand side of the eqs. (33) equal to zero
and solving the resulting system for r a and rb to give
(34)
The configuration of two collateral vessels is feasible provided the solution (34) represents
a stable stationary state. To investigate the stability of the solution obtained it is necessary to
study the solution of the system (33) for initial conditions representing small perturbations of
the radii around the values given by eq. (34). Keenen and Rodbard (1973) solved the non-linear
system of governing equations numerically using arbitrary values of the parameters not related
to a specific bifurcation or bypass. As initial conditions at t = 0 one of the radii was taken to
be slightly bigger than the stationary solution r*, and the other was chosen to be slightly smaller
than r*. The solution obtained showed that over time the smaller radius diminishes to zero, while
the bigger radius tends to r c = ~' which yields the baseline value of the wall shear stress
for total flow Q. The authors drew the conclusion that the configuration of two collateral vessels
is unstable and therefore the flow-induced remodeling modulated by the shear stress is accepted
as the probable mechanism of the physiological observations. Therefore reduction of the number
of collateral arteries and the total occlusion of the bypassed artery is considered to be an adaptive
response of the arterial wall aiming to maintain an optimal shear stress level.
The qualitative theory of dynamical systems allows the study of stability of their stationary
solutions without solving the equations. The basic idea is to assess the character of the solution
of the linearized system of equations that describes small perturbations around the stationary
solution. The type and signs of the eigenvalues of the linearized system, which in tum depend
on its coefficients, determine whether the stationary solution obtained is stable or not. It can be
shown that the stationary solution (34) is always unstable and it represents a saddle point on the
phase portrait of the dynamic system (33). Similar stability analyses were performed by Harrigan
(1997), who studied the feasibility of several simplified models of shear stress-sensitive vessels
connected in parallel and in series with passive resistances, as representative of configurations of
arterioles in the vascular network.
4.3 Stress-Induced Geometrical Remodeling of Vessel Segments Adjacent to Stents and

Artery/Graft Anastomoses
Experimental studies have shown that in the zone adjacent to an implanted Palmaz-Schatz stent
the arterial lumen decreases due to wall remodeling. The wall thickening gradually decreases
with distance from stent edges (Hoffman et al. (1996)). Marked thickening also occurs at the
artery/graft anastomoses, especiallywhen low compliance synthetic grafts were used (DeWeese
(1985) and Echave et al. (1979)). On the other hand, theoretical studies reveal that marked in-
creases in axial and circumferential stresses appear in zones close to end-to-end anastomosis of
a graft and a host artery (Paasche et al. (1973), Chandran et al. (1992) and Matsumoto et al.
(1994)).
It was shown in the previous sections of this paper that arteries change their geometry in
response to alterations in their mechanical environment tending to restore the baseline values of
flow-induced shear stress and pressure-induced wall tensile stress. It is reasonable to expect that
observed changes in the arterial thickness in a host artery close to a stent or graft represents an
adaptive response to the stress concentration. However, if remodeling is considered to be driven
exclusively by wall stresses there is no a priori reason to expect the process to be self-limiting. A
theoretical study addressed to this issue was performed by Rachev et al. (2000).
Mathematical model of arterial wall. Firstly, the artery was considered as a long cylindrical
shell of constant mid-wall radius and constant wall thickness. The shell was considered clamped
at the junction to a graft or stented arterial region. The wall material was assumed to be elastic,
incompressible and orthotropic with axes of orthotropy in the axial and circumferential direction.
The artery has extended longitudinally to its in situ length and was subjected to variable inter-
nal pressure. Given the loads, initial dimensions, and prescribed boundary conditions the wall
stresses were calculated. It was hypothesized that when the magnitudes of the stresses exceed
certain threshold values the wall thickens according to a postulated remodeling rate equation and
the vessel becomes a shell of variable thickness.
To calculate the stresses that exist in the arterial wall at an arbitrary moment the deforma-
tion process was divided into two stages. First, the artery was considered apart from the graft or
stented region as a cylindrical membrane of constant mid-wall radius and ofvariab1e wall thick-
ness. The membrane is extended to its in situ length and is inflated by the mean arterial pressure
undergoing a finite axisymmetric deformation. The membrane stresses am and their resultants
N m are calculated as in Appendix B (Figure 19). The subscripts 1 and 2 refer to the axial and the
circumferential direction, respectively, and XI is the axial coordinate measured from the clamped
cross-section along the deformed membrane.
The real artery is clamped at the anastomosis and is subjected to pressure that varies from
diastolic to systolic value. Therefore, after being finitely extended the artery was subjected to
additional axisymmetric deformation caused by clamping and by a pressure that is a difference
between the variable and mean pressure. In this stage, the vessel was considered to be a long
prestressed cylindrical shell subject to inflation and bending. The middle surface undergoes an
additional displacement with component w (XI) in the radial direction, which is considered small
compared with the wall thickness. The displacement in the axial direction was assumed to be
zero, in agreement with the experimental observations for the tethering effect of the tissues sur-
rounding the artery during pressure pulsation. Stress resultants due to the bending deformation
are additional axial and circumferential tensile tensions, an axial transverse force and axial and
254 A. Rachev
....
Figure 19. Schematic representation of the membrane state.
circumferential bending moments. They are denoted by prime and are shown in Figure 20.
Following the linear theory for bending of cylindrical shells the governing equations for the
deformation process were reduced to the following linear partial differential equation for the
radial displacement w,
(35)
where
(36)
h(x 1 ) is the wall thickness and r is the mid-wall radius resulting from the finite membrane defor-
mation; E 1 , E 2 , f..Ll and f..L 2 are the incremental elastic moduli and coefficients of the transversal
deformation, which characterize the linearized mechanical behavior of the shell around its mem-
brane deformed state (Rachev et al. (2000)); P is the variable pressure, which changes from
Pd - P to P8 - P during the cardiac cycle; P8 and Pd are the systolic and diastolic pressure and
P is the mean arterial pressure.
Because the shell is clamped at x 1 = 0, this imposes the boundary conditions
dw _ 0
w = R-r, dx1 - '
(37)
where R is the radius of the graft or stent at the anastomosis.

Stresses due to bending are linearly distributed across the wall thickness and have their ex-
treme values at the inner and the outer wall surfaces. The total stresses are the sums of the stresses
due to the finite membrane deformation and stresses due to the additional bending deformation.
Far away from the anastomosis the artery does not experience bending and the stress field corre-
sponds to the membrane deformed state.
i
~,h1 + N2
I
l
/ /
/
------~--~--~
/
Figure 20. Schematic representation of the bending state.
Remodeling rate equation. Equation (3 5) contains two unknown functions, w (x 1 ) and h( x 1 ).

Another differential equation for w(xl) and h(xi) was derived assuming a stress-induced wall
thickening of the artery. It was postulated that the rate of change of the wall thickness is propor-
tional to the deviation of the current circumferential and axial stresses from their values existing
far away from the clamped edge and caused by the systolic pressure, as follows
where (at (x1 , t)) and (a2 (x1, t)) are the axial and the circumferential stress, averaged across
the thickness and over one heart cycle (see Rachev et al. (2000)); af and a2 are the membrane
stresses corresponding to systolic pressure, a :::; 1 is a weighting coefficient, which accounts
for the different contribution of the axial and circumferential stresses to the remodeling process;
k :::; 1 is a coefficient that specifies the threshold values of the stress increments which initiate
remodeling; T is a time constant, and h 0 is the membrane thickness far from the clamped cross-
section.
Results and Discussion. To obtain theoretical results from the proposed mathematical model,
the model parameters were specified for a canine carotid artery and the governing equations were
solved numerically using a finite difference method and an explicit time step.
The predicted time course of the wall thickness at the clamped region (x 1 = 0) from the
onset of stent/graft implantation is shown in Figure 21 for different values of the parameter k. A
similar monotonic behavior in the time variation of the thickness was noted at other locations.
Figure 22 illustrates the predicted variation of the wall thickness in the final adapted state for
different values of the parameter k. It is seen that the wall thickening is localized in the zone close
to anastomosis. Wall thickness gradually decreases with distance from the edge of the stent/graft
256 A. Rachev
2.5
k=l.2
-
~
...._
~
'-' 2.0 k=l.3

...c:
"'v"' k=1.4
j
-
(.)
:E 1.5
"0
v
-~
(ij
§ 1.0
z0
0.5
0 100 200 300
Dimensionless time t/T
Figure 21. Predicted time course of the normalized wall thickness at the clamped region (x 1 = 0) from the
onset of stent/graft implantation for different values of the parameter k.
to the 'normal' value far from the treatment site. These results agree with the reported experi-
mental findings (Hoffman et al. (1996), De Weese (1985) and Echave et al. ( 1979)).
The axial stress variation along the vessel in the final adapted state compared to the stresses
existing at the onset of graft or stent deployment is given in Figure 23. As expected wall thicken-
ing reduced the stress concentration in zones adjacent to the clamped cross-section. Therefore, if
the stress concentration is accepted as a major determinant of vessel wall remodeling, this model
predicted that the process is self-limiting and leads to local thickening of the host artery in the
vicinity of an implanted stent or graft.
Despite the predicted stability of the remodeling process, in the case of pronounced increase
in wall thickness, some assumptions on which the model is based might not remain valid and
need improvement. Consideration of an artery as a thin shell requires the deformed wall thick-
ness to be significantly less than the mid-wall radius. In the general case the strain and stress
analysis have to be performed using appropriately developed FEM for finite nonlinear elastic
deformations.
The model assumed that the increase in thickness was symmetric about the mid-wall sur-
face, but this may not be the case in vivo. Determination of the real distribution of the new
material across the arterial cross-section requires more detailed experimental investigations than
have been performed previously. Mechanical properties and muscular tone were considered un-
changed during remodeling and the arterial wall was considered as one-layered shell made of
homogeneous material. Experimentally observed by some authors (Hasson et al. (1985)) hyper-
compliant zones near vascular anastomosis might result from stress-induced changes in the wall
material. A theoretical test of such a hypothesis requires consideration of an artery of variable
0.4
*
~ 0.3
~
"'
]"'u
:§ 0.2
'"0
(])
.!::3
a
~
i 0.1
0 0.2 0.4 0.6 0.8
Normalized length xdr*
Figure 22. Predicted variation of the deformed wall thickness in the final adapted state for different values
of the parameter k.
mechanical properties in the axial direction.

Finally, if stress-induced wall thickening is very pronounced, the effects of an altered flow
pattern near a junction might play a significant role in arterial remodeling and cannot be disre-
garded. In fact, stent and graft failure, including restenosis and lumen obliteration, are likely due
to combined effects of abnormal flow-induced shear stresses at the intima and tensile stresses in
the arterial wall. The study was directed, however, not to simulate conditions and consequences
which lead to stent or graft failure, but rather to find the domain of model parameters for which
abnormal wall thickening does not occur or is manifest within admissible limits. However, the
appropriate choice of and motivation for the remodeling rate equation remains a key issue for
further investigations.
4.4 A Model of Arterial Regeneration Over Bioresorbable Grafts
Unsatisfactory results in using small caliber synthetic grafts motivate investigations aimed at im-
proving their patency rates. There exists an increasing interest in the use ofbioresorbable material
for fabricating vascular prostheses. They are made of absorbable lactide/glicolide copolymers.
The graft dissolves due to hydrolysis and enzymatic degradation over post-implant time forming
a scaffold for tissue ingrowth that leads to development of a neo-vascular conduit. In the early
implanted phase, which last from one to two weeks, the graft begins to resorb, becoming thinner
and softer. Simultaneously a thin and soft inner and an outer layer develop. Over time these lay-
ers, and predominantly the inner one, thicken and the material gradually increases its stiffness.
The rate of ingrowth parallels the kinetics of the resorption of the graft. After approximately two
to three months the graft is practically resorbed. The growth and strengthening of the arterial
tissue continue in a phase of vessel maturation until the process of formation of a neo-arterial
258 A. Rachev
800
600
~
~
"'
'-' 400
"'
...."'.....
Q)
~
"' 200
.R
<
0
-200
0 2 3
Normalized length xdr•
Figure 23. Predicted variation of the axial stress at the inner surface and outer surface at diastolic pressure
Ps = 60 mmHg. k = 1.2. Continuous curves refer for the moment immediately after stentor
graft implantation; dotted curves refer to the final adapted state. The horizontal line corresponds to
the stress existing before implantation of a stentor a graft for systolic pressure Ps = 140 mmHg.
conduit is complete. Figure 24 shows the experimentally recorded time course of the thickness
of the inner neo-artery layer when different resorbable grafts were implanted.
The interrelation between the resorption process and factors driving the arterial regeneration
are poorly understood. To the best of our knowledge, there exists only one theoretical investi-
gation devoted to analysis of the stress distribution in the resorbable graft/neo-artery complex
(Vorp et a!. (1995)). The authors used the classical Lame solution for a three-layered tube in-
flated by a uniform pressure to study the contribution of the thickness and modulus of the graft
and neo-artery to the stress distribution. The major result was that early in the resorption phase
the neo-artery is subjected to compressive circumferential stresses, suggesting the hypothesis
that they promote neo-artery formation. No coupling between the dynamics of the neo-artery
formation and the strain and stress states in the vessel wall was considered.
During formation of a neo-artery the pressure and flow remain practically unchanged. How-
ever, a decrease in graft thickness and softening of the graft material causes a redistribution of
the load in the artery/graft complex, and therefore the stresses and strains experienced by the
neo-arterial tissue and sensed by the smooth muscle cells vary in time. It is reasonable to assume
that the mechanisms of neo-arterial formation, at least in part, are similar to those of arterial wall
remodeling in response to mechanical factors. To test this hypothesis and to estimate the effects
of certain mechanical and geometrical graft parameters on the neo-artery formation a theoretical
model has been proposed (Rachev and Kirilova (2001) and Kirilova and Rachev (2002)).
Mathematical model. The neo-artery/graft complex was considered to be a thick-walled tube.

At the moment immediately after graft implantation it consists of one layer representing the
bioresorbable graft (Figure 25a). During graft resorption and neo-artery formation the tube com-
prises two layers (Figure 25b ). Finally, after the graft is completely resorbed but the neo-artery
800 -
600 -
400 -
200 -
2 3
Months
Figure 24. Time course of inner layer thickness for different graft material (Greisler eta!. (1987)).
formation is still in progress the tube is considered again as a single layer (Figure 25c ). Both the
artery and graft material are assumed to be elastic incompressible and orthotropic. As might be
expected, and has been shown by the linear analysis ofVorp et al. (1995), accounting for an outer
neo-arteriallayer has no significant contribution to the stress redistribution. Under applied load
the tube undergoes an axisymmetric finite deformation and is in a state of plane strain. Arterial
pressure and blood flow rate are considered to be constant during the entire process of neo-artery
formation.
The stress and strain analysis of the neo-artery/graft complex follows the method described
in Appendix A. The strain-energy density functions for the material of the neo-artery and the
graft were chosen to be quadratic functions of the Green strains (Patel and Vaishnav (1972) and
Stewart and Lyman (1988)),
W = Ae~ + Beoez + Ce;, (39)
where A, B and C are material constants different for the artery and the graft. Due to lack of
experimental data for the manner in which material constants A, B and C vary during the process
of neo-artery formation and graft resorption, it was accepted that at any moment the constants B
and C change in proportion to A.
For simplicity, the constant Ag of the graft material was assumed to decrease in a linear
manner during post-implant time,
Ag(t) = A~(l- at), (40)
where A~ is the value at the moment of graft implantation, and a is a constant characterizing the
decay rate.
260 A. Rachev
(a) (b) (c)
Resorbable graft Resorbable graft Neo-artery
Figure 25. Schematic representation of the cross-sections of(a) the graft immediately after implantation; (b)
the neo-artery/graft complex; and (c) the neo-artery during maturation.
It was assumed that the inner radius of the neo-artery/graft complex remains constant. The
!
neo-artery regenerates growing outwards according to the remodeling rate equation
T1 (aa-ae).tfae < aa,

dH ab
d/ =
1
Oifaa < ae < aa, (41)
~2 ( ae ~ ab) if ae > ab,
where Ha is the undeformed arterial thickness; ao is the current mean circumferental arterial
stress; a a and ab are the 'equilibrium' physiological stresses; T 1 and T2 are time constants that
depend on the graft material. Equation (41) states that the rate of change of H a depends linearly
on the deviation of the current mean arterial stress from certain values, which do not elicit arte-
rial remodeling. In that respect eq. (41) is consistent to some extent with the phenomenological
stress-growth law proposed by Fung and shown schematically in Figure 6.
It is assumed that the kinetics of the regenerative thickening parallels the time course of graft
dissolution, so that
Ha(t) = H~- Hg(t), (42)
where H~ and Hg(t) are the initial and current graft thickness, respectively.
Finally, the evolution equation for the material constant Aa is postulated in the form
dA ~ ~~ (ktC~o-C) ifC :S ktCo,

dt = OifktCo < C < k 2 C0 , (43)
_!_
T2
(c- k2Co)
Co t
·fc >-
k C
2 o,
where C is the current compliance of the artery/graft complex calculated usingthe mid-wall ra-
dius of the neo-artery; Co is the baseline value of the arterial compliance; kt < 1 and k2 > 1
are coefficients that specify the threshold values of the compliance mismatch, which can elicit
changes in the mechanical properties of the neo-artery; T1 and T2 are time constants. Accord-
ing to eq. (43) the arterial material increases its stiffness in response to an abnormal pulsatile
stretch of the arterial wall estimated through the deviation of the current arterial compliance
from its baseline value. Equation (43) is in agreement with the experimental observations that
cyclic overstretching as well as loss of stretch experienced by the smooth muscle cells provokes
their synthetic and proliferative activity (Lehoux and Tedgui (1998)).
Thus, formation of a neo-artery is described by a system of two non-linear non-autonomous
first-order differential equations (41) and (43) for evolution of the arterial geometry and mechan-
ical properties. The equations are coupled to the equation of equilibrium of the vessel and to the
equations describing the changes in the geometry and mechanical properties of the resorbable
prosthesis.
Results and Discussion. The dimensions and the mechanical properties of vessel material
were specified for a rabbit aorta and a resorbable graft using data in (Patel and Vaishnav (1972)
and Stewart and Lyman (1988)). The model parameters and time constants were chosen to yield
a reasonable time course of the regeneration and resorption processes. The governing equations
were solved numerically. The material constants of the graft material, the stress distribution and
the compliance of the artery/graft complex were calculated at each time step. The value of the
mean arterial stress and arterial compliance were substituted into the evolution equations and the
new thickness and elastic constants of neo-artery were calculated. The procedure was repeated
and terminated, provided it is self-limiting, when the graft thickness vanished and the dimensions
and mechanical properties of the neo-artery attained values close to those of a normal artery.
The time course of the arterial thickness exhibits anN-shaped pattern (Figure 26a), similar
to the experimental observation of Greisler et al. (1987) for regeneration of a rabbit aorta over
a bioresorbable PGA (polyglicolic acid) graft (Figure 24). Material constants of arterial tissue
increase monotonically to attain the values of a normal artery (Figure 26b ), while the graft con-
stants decrease following the prescribed linear dependence on time (not shown).
The stress experienced by the artery (Figure 26c) is compressive in the early stage of graft
resorption, as was found in the linear study ofVorp et al. (1995). A decrease in stress after com-
plete resorption of the graft is due to progressive thickening of the artery, which normalizes the
wall stress. Similar is the pattern of the compliance, which drives the stiffening of the neo-arterial
material (Figure 26d).
A parametric study was performed varying the thickness, stiffness and rate of mechanical
softening of the bioresorbable grafts. It was shown that a fast resorption of graft material leads
to overloading of the neo-artery tissue before it is sufficiently organized to bear load. This might
lead to bursting or aneurismal dilatation, as had been observed experimentally (Greisler et al.
(1987) and Greisler et al. (1988)). Similar effects are predicted theoretically when the graft ex-
hibits a small initial wall thickness.
In conclusion, the results obtained supported the hypothesis that arterial regeneration over a
bioresorbable graft is an adaptive process that is influenced and modulated by mechanical fac-
tors. If the mean circumferential stress and pulsatile stretch are accepted as major determinants
of the vessel wall regeneration, the process is self-limiting and leads to formation of a vessel with
dimensions and mechanical properties close to those of a normal artery. More slowly absorbed
and/or thicker prostheses might more reliably allow for adequate strength necessary to prevent
262 A. Rachev
(a) (b)
1.0 30
s5 0.8
~
"'
1'.
0
.,:; 20
"'
_,."""' <:
~
.::l 0.5
-s "
0
c
u
c
"'l' 0.3
t:: "t::
10
0 "'6
z" z"
0
0.0
0 2 3 0 1 2 3
Time (months) Time (months)
(c) (d)
200 0.04
-;;-
1'.
0 150 0.03
"'"'
~ "'
1'.
0
""' 100
" 0.02
"a u
c "
.~
P..
~" a0
50
0.01
6 u
z"
0
0.00
0 I 2 3 0 2 3
Time (months) Time (months)
Figure 26. Predicted time course of: (a) neo-artery wall thickness; (b) neo-artery material constant Aa; (c)
mean circumferential stress in the neo-arterial wall; (d) compliance of the neo-artery/graft com-
plex.
bursting or aneurysmal dilatation.

Despite realistic model predictions more experimental studies are needed to verify the hy-
potheses introduced and to identify the model parameters and functions. The model suggests the
need for quantifying the time course ofthe radius and wall thickness of neo-artery and resorbable
graft and the time course of their mechanical properties and relevant constitutive formulation of
these processes. After experimental validation of the proposed remodeling rate equations by
comparison of the theoretical results with experimental findings, model predictions can be used
to improve graft design and selection.
Appendices
Use of solid mechanics for a strain and stress analysis of an artery requires adoption of a relevant
mathematical model. Modeling includes: i) assumptions about the arterial geometry (geometrical
model); ii) assumptions about the interaction of the vessel with other bodies (model of applied
loads and boundary conditions); iii) mathematical description of the mechanical properties of the
(a)
Figure27. Schematic representation of the arterial cross-section (a) at the zero-stress state and (b) at the
current defonned state.
vascular tissue (selection of constitutive equations); and iv) modeling of the deformation process
by introducing hypotheses for the character of the strain and/or stress state in the arterial wall.
Depending on geometrical dimensions of a specific artery and on the objectives of the investiga-
tion, the blood vessel might be considered as a three-dimensional or as a two-dimensional body.
Appendix A
A 3-D Stress and Strain Analysis of Arteries
As have been demonstrated in several studies (Vaishnav and Vossoughi (1983) and
Chuong and Fung (1986)) a radial cut of an unloaded arterial segment makes it spring open
and the cross-section takes a shape close to that of a circular sector (Figure 27a). Assuming that
in this configuration the vessel is in the zero-stresses state, the cut-opened configuration is used
as a reference configuration to define strains. The angle Cb, called the opening angle, was intro-
duced by Chuong and Fung. It is an indicator of the existence of residual stresses and strains in
the state of no load. Under physiological conditions an artery is subjected to an internal pressure
and is extended in the axial direction. It undergoes an axisymmetric finite deformation from the
reference to the deformed configuration, which is considered to be a circular thick-walled tube
in the state of plane strain (Figure 27b).
The arterial tissue is assumed to be an orthotropic incompressible elastic material, with axes
of orthotropy in the radial, circumferential and longitudinal directions. The constitutive equations
follow from a strain-energy density function (SEF) that is a function of the principal strains.
Given the dimensions at the zero stress state, SEF, the internal pressure, and the axial stretch
ratio (or the total axial force), the strain and stress distribution in the arterial wall is determined
using a semi-inverse method. The deformation from the zero-stress state to the current deformed
state is prescribed in terms of relations that contain a finite number of unknown parameters. After
having calculated the strain and stress fields, the parameters were determined from the equations
of equilibrium and the boundary conditions.
264 A. Rachev
Using notations given in Figure 27, the deformation is described by the relations
7r
X= --,n-, (Al)
Jr-~
where JL, x and .X are deformation parameters; L and l are the lengths of the arterial segment in
the zero-stress state and the deformed state, respectively.
Following the theory of finite deformations (Green and Adkins (1970)) and denoting the
cylindrical coordinates of an arbitrary point before and after deformation by (R, 8, Z) and
(r, 0, z), the stretch ratios in the radial, circumferential and axial direction are
7rr xr (A2)
.Ae = (1r- P)R R'
The principal components of the Green strain are
1 2
ei = 2(\ -1), i = r,O,z. (A3)
It follows for the incompressibility of the wall material that ArAIJAz = 1 and hence
(A4)
After integrating eq. (A4)
r= (A5)
The strain-energy function can be given in two alternative forms. If W W(er, ee, ez) the
constitutive relations between the principal Cauchy stress and Green strains are
(A6)
where p is an unknown scalar function, which appears due to the material incompressibility.
If the condition of incompressibility is used to eliminate the radial strain er then W
W(ee, ez) and the constitutive equations are
(A7)
Both the formulations of the constitutive equations given by eqs. (A6) and (A7) are three-
dimensional.
Equations of equilibrium yield
dur O"r - O"IJ 8p

-dr+ r
=0, ao = o, 8p = 0
8z '
(A8)
and the boundary conditions at the inner and outer surface are u r ( r = ri) = - P, u r ( r = r 0 ) =
0, where P is the arterial pressure. Integrating eq. (A8) and using the first boundary condition
(A9)
Given the opening angle the parameter x is known. The remaining deformation parameters
1-l and A are to be determined from prescribed external load as follows
(AlO)
F is the total axial force acting on the arterial segment. If the axial stretch ratio A instead of
force F is given, the only unknown parameter 1-l is calculated using eq. (A10)1. After having
determined the deformation parameters, the strain and stress distributions in the arterial wall are
calculated using eqs. (AS), (A2), (A4) and (A9). When the strain-energy function is given in the
form W = W(ee, ez) the eqs. (A9) and (A10) are modified accordingly by setting the derivatives
with respect to er equal to zero and replacing W by W.
In the particular case of an unloaded arterial segment the unknown deformation parameters
1-l and A are to be determined from the eqs. (AlO) by setting P = 0 and F = 0. Calculations
have shown that with good accuracy the value of the deformation parameter A is close to 1, as
was set by Chuong and Fung (1986). The strains and stresses corresponding to the state of no
load are called the residual strains and stresses. Because no load is applied, the residual stresses
are equilibrated in the artery wall. They are compressive in a portion of the wall thickness close
to the inner surface and are tensile in a portion of the wall close to the outer surface. When a
thick-walled tube is subjected to an internal pressure the circumferential stress is higher at the
inner surface. Existence of residual strains reduces the strain gradient across the arterial wall un-
der physiological load conditions and leads to optimal, from a mechanical point of view, bearing
of applied load by the structural elements of the media (Chuong and Fung (1986)). Moreover,
a close to uniform distribution of the circumferential strain and stress provides a uniform local
mechanical environment for vascular smooth muscle cells throughout the arterial wall and thus
optimizes their performance. Evidently there exist no residual strains if the opening angle is zero.
The analysis given above was generalized for the case when an artery was considered to be a
two-layered tube (for details see Rachev (1997)). The field equations were written for each layer
and conditions for continuity of the radial displacement and the radial stress were imposed. Given
that the axial stretch ratio is the same in both layers, the stress and strain fields again depend on
a single deformation parameter, which is determined form an equation similar to eq. (A1 Oh but
the integration is performed over the thickness of both the layers.
To account for the contribution of the vascular smooth muscle Rachev and Hayashi (1999)
modified the given analysis in two directions. Firstly, the opened-up configuration when the
smooth muscle cells are maximally relaxed was taken to be a zero-stress reference state. Sec-
ondly, the circumferential stress was represented as the sum of a stress calculated from the con-
stitutive eqs. (A 7) or (A8), and termed as passive stress, and an active circumferential stress
developed by the stimulated smooth muscle (see eq. (22)).
266 A. Rachev
(a) (b)
Figure 28. Schematic representation of the arterial cross-section (a) at the zero-stress state and (b) at the
current deformed state.
Appendix B
A 2D-Stress and Strain Analysis of Arteries
In some cases an artery can be considered as a thin-walled circular cylindrical membrane of
constant thickness. Its deformed state is described by the deformation of the membrane mid-wall
surface. The stresses acting in cross-sections normal to the middle surface are assumed uniformly
distributed across the membrane thickness, while the stresses in cross-sections parallel to the
middle surface are assumed to be zero. Stress and strain analysis based on the three-dimensional
consideration has shown that the membrane approach is asymptotically valid provided the fol-
lowing conditions hold true: i) the thickness is sufficiently smaller than the other dimensions
of the membrane; in particular, for a circular cylindrical membrane the thickness to radius ratio
has to be less than 0.2; ii) the radius, thickness, mechanical properties and applied load vary in
a smooth manner along the membrane surface; iii) the boundary conditions do not impose re-
strictions on the deformations of the middle surface. These conditions are satisfied for healthy
non-stenosed aorta and large arteries subjected to physiological loads and located far way from
bifurcations.
The state of no load is taken as a reference state for calculating the strain components (Fig-
ure 28a).
When the membrane is inflated by an internal pressure and is extended longitudinally it un-
dergoes an axisymmetric finite deformation (Figure 28b ). The membrane stretch ratios in the
axial and circumferential directions are
(Bl)
where Rm and rm are the mid-wall radii of the artery at the unloaded and deformed state re-
spectively, and L and l are the corresponding axial lengths. It follows from the condition of
incompressibility that
l H
(B2)
Ar = Az .Ae ' h = AzAe'
' '
Figure 29. Free body diagram of an artery.
where h and H are the deformed and undeformed wall thickness, respectively. The Green strains
are again given by eq. (A3).
The membrane circumferential and axial stresses are uniformly distributed across the wall
thickness, while the radial stress is zero everywhere. Depending on the form in which the mem-
brane strain-energy density function is given the constitutive equations are
(B3)
if the strain-energy function is given in the form Wm = Wm(er, eo, ez) and
(B4)
if the strain-energy function is given in the form Wm = Wm (eo, e z). Both the formulations are
two-dimensional. It is worth noting that the membrane strain-energy functions W m and Wm are
functions of the strains calculated for the middle surface, and are in general different from the
three-dimensional strain-energy functions W and W.
Because the stresses are uniformly distributed across the wall thickness their stress resultants,
called the membrane tensions, are
(B5)
The equation of overall equilibrium in the radial direction (Figure 29) yields a 0
formula known as the law ofLaplace. Using eqs. (Bl) to (B3) it takes the form
ao(>..o) - P 1)
( -AzHA~- R - -2 = 0
. (B6)
Given the zero-stress configuration, the strain-energy function, the axial stretch ratio Az and
the applied pressure, the unknown stretch ratio >..o is determined from eq. (B6) and all membrane
stresses and strains can be calculated. It is evident from of eq. (B6) that the membrane approach
implies that the unloaded membrane is in the stress-free state.
The membrane approach is simpler than the more general 3-D consideration. However, it
does not allow determination of the strain and stress distribution across the wall thickness and
268 A. Rachev
disregards the existence of radial stress. Moreover, the membrane consideration neglects the ef-
fects of the residual strains. Therefore, it excludes a comparison between theoretically predicted
values of the opening angle and its experimental records to be used as a tool for validating mod-
els of arterial remodeling and growth.
A systematic and comprehensive description of arterial wall mechanics and a comparative
study of most material models is given by Holzapfel et al. (2000).
Acknowledgments
The author gratefully acknowledges the help of Prof. R.W. Ogden and Prof. G.A. Holzapfel in
editing the manuscript.
References
Achakri, H., Rachev, A., Stergiopulos, N., and Meister, J. J. (1994). A theoretical investigation of low
frequency diameter oscillations of muscular arteries. Ann. Biomed. Engr. 22:253-263.
Albridge, H. E., and Trimble, A. S. (1971). Progression of coronary artery lesions to total occlusion after
aorta-coronary saphenous vein bypass grafting. J. Thorac. Cardiovasc. Surg. 6:7-11.
Berry, C. L., and Greenwald, S. E. (1976). Effects ofhypertension on the static mechanical properties and
chemical composition of rat aorta. Cardiovasc. Res. 10:437--451.
Brownlee, R. D., and Langille, B. L. (1991). Arterial adaptation to altered blood flow. Canad. J. Physiol.
Pharm. 69:978-983.
Chandran, K. B., Gao, D., Han, G., Baraniewski, H., and Corson, J.D. (1992). Finite-element analysis of
arterial anastomoses with vein, dacron and PTFE grafts. Med. Bioi. Eng. Comput. 30:413--418.
Chuong, C. J., and Fung, Y. C. (1986). On residual stress in arteries. J. Biomech. Engr. 108:189-192.
Cox, R. H. (1978). Regional variation of series elasticity in canine arterial smooth muscle. Am. J. Physiol.
234:542-551.
DeWeese, J. A. (1985). Anastomotic neointimal fibrous hyperplasia. In Brenhard, V. M., and Towne, J. B.,
eds., Complications in vascular surgery, 157-170. Orlando: Grune and Stratton.
Dobrin, P. B. ( 1973). Influence of initial length on length-tension relationship of vascular smooth muscle.
Am. J. Physiol. 225:659-663.
Echave, V., Koomick, A. R., Haimov, M., and Jacobson, J. H. (1979). Intimal hyperplasia as a complication
of the use of the polytetrafiuoroethylene graft for femoral-popliteal bypass. Surgery 86:791-798.
Fath, W., Burkhart, H. M., Miller, S.C., Dalshing, M. C., and Unthank, J. L. (1998). Wall remodeling after
wall shear rate normalization in rat mesenteric collaterals. J. Vase. Res. 35:257-264.
Fridez, P., Makino, A., Miyazaki, H., Meister, J. J., Hayashi, K., and Stergiopulos, N. (200la). Short-term
biomechanical adaptation of the rat carotid to acute hypertension: contribution of smooth muscle. Ann.
Biomed. Engr. 29(1):26--34.
Fridez, P., Rachev, A., Meister, J. J., Hayashi, K., and Stergiopulos, N. (200lb). Model ofgeometrcal and
smooth muscle tone adaptation of carotid artery to step change in pressure. Am. J. Physiol. Heart Circ.
Physiol. 280:H2752-H2760.
Fung, Y. C., and Liu, S. Q. (1989). Change of residual strains in arteries due to hypertrophy caused by
aortic constriction. Circ. Res. 65:1340-1349.
Fung, Y. C., and Liu, S. Q. (1991). Change of zero-stress state of rat pulmonary arteries in hypoxic
hypertension. J. Appl. Physiol. 70:2455-2470.
Fung, Y. C., Liu, S. Q., and Zhou, J. B. (1993). Remodeling of the constitutive equation while a blood
vessel remodels itself under stress. J. Biomech. Engr. 115:453--459.
Fung, Y. C. (1990). Biomechanics. Motion, Flow, Stress and Growth. New York: Springer-Verlag.
Fung, Y. C. (1991). What are the residual stresses doing in our blood vessels? Ann. Biomed. Engr. 19:237-
249.
Fung, Y. C. (1996). Biomechanics. Circultion, 2nd ed. New York Berlin Heidelberg: Springer-Verlag.
Glagov, S., Zarnis, C. K., Masawa, N., Xu, C. P., Bassiouny, H., and Giddens, D.P. (1993). Mechanical
functional role of non-atherosclerotic intimal thickening. Frontiers Med. Bioi. Eng. 1:37--43.
Green, A. E., and Adkins, J. E. (1970). Large Elastic Deformations and Nonlinear Continuum Mechanics.
London: Oxford University Press.
Greisler, H., Ellinger, J., Schwarcz, T., Golan, A., Raymond, R., and Kim, D. (1987). Arterial regeneration
over polydioxanone prostheses in the rabbit. Arch. Surg 122:715-721.
Greisler, H., Endean, E., Klosak, J., Ellinger, J., Dennis, J., Buttle, K., and Kim, D. (1988).
Polyglactin910/polydioxanone bicomponent totally resorbable vascular prostheses. J. Vase. Surg.
7:697-705.
Han, H.-C., and Ku, D. N. (2001). Contractile response of arteries subjected to hyperensive pressure in
seven-day organ culture. Ann. Biomed. Engr. 29:467--475.
Han, H.-C., Vito, R. P., Micheal, K., and Ku, D. N. (2000). Axial stretch increases cell proliferation in
arteries in organ culture. Advances in Bioengineering, ASME 48:63-64.
Harrigan, T. P. (1997). Regulatory interction between myogenic and shear-sensitive arterial segments:
conditions for stable steady states. Ann. Biomed. Engr. 25:635-643.
Hasson, J. E., Megerman, J., and Abbott, W. M. (1985). Increased compliance near vascular anastomoses.
J. Vase. Surg. 2:419--423.
Hayashi, K., Makino, A., Kakoi, D., and Miyazaki, H. (2001). Remodeling of arterial wall in response
to blood pressure and blood flow changes. Proceeding of 2001 Summer Bioengineering Conference
819-820.
Hayashi, K. (2000). Private communication.
Hill, A. V. (1939). The heat of shortening and the dynamic constants of muscle. Proc. Roy. Soc. London
Ser. B 126:136-195.
Hoffman, R., Mintz, G. S., Dussaillant, G. R., Popma, J. J., Pichard, A. D., Satler, L. F., Kent, K. M.,
Griffinn, J., and Leon, M. B. (1996). Patterns and mechanisms ofin-stent restenosis. a serial intravas-
cular ultrasound study. Circ. 94:1247-1254.
Holzapfel, G. A., Gasser, T. C., and Ogden, R. W. (2000). A new constitutive framework for arterial wall
mechanics and a comparative study of material models. J. Elasticity 61:1--48.
Hutchins, G. M., Miner, M. M., and Boitnott, J. K. (1976). Vessel caliber and branch-angle of human
coronary artery branch-points. Circ. Res. 38:572-576.
Kamiya, A., and Ando, J. (1996). Response of vascular endothelial cells to fluid shear stresses: Mechanism.
In Hayashi, K., Kamiya, A., and Ono, K., eds., Biomechanics, Functional Adaptation and Remodeling,
9-10. Tokyo: Springer-Verlag.
Kamiya, A., and Togawa, T. (1980). Adaptive regulation of wall shear stress to flow change in canine
carotid artery. Am. J. Phys. 239: 14-21.
Keenen, R., and Rodbard, S. (1973). Competition between collateral vessels. Cardiovasc. Res. 7:670-675.
Kirilova, M., and Rachev, A. (2002). A theoretical investigation of the effects ofbioresorbable material on
process ofneo-artery regeneration. Theoretical and Applied Mechanics (Bulgaria). in press.
Langille, B. L., and O'Donnell, F. (1986). Reductions in arterial diameter produced by chronic decrease in
blood flow are endothelium-dependent. Science Wash. DC 231:405--407.
Langille, B. L., Bendeck, M. P., and Keeley, W. ( 1989). Adaptations of carotid arteries of young and mature
rabbits to reduced carotid blood flow. Am. J. Physiol. 256:931-939.
Langille, B. L. (1993). Remodeling of developing and mature arteries: Endothelium, smooth muscle and
matrix. J. Cardiovasc. Pharmacol. 21:11-17, (Suppl. 1).
Langille, B. L. (1995). Blood flow-induced remodeling of the artery wall. In Bevan, J. A., Kaley, G.,
and Rubanyi, G., eds., Flow-dependent regulation of vascular function, 227-299. New York, Oxford:
American Society of Mechanical Engineers.
270 A. Rachev
Lehoux, S., and Tedgui, A. (1998). Signal transduction of mechanical stresses in the vascular wall. Hyper-
tension 32:338-345.
Leung, D. Y. M., Glagov, S. G., and Mathews, M. B. (1976). Cycling stretching stimulates synthesis of
components by arterial smooth muscle cells in vivo. Science 116:455-477.
Liu, S. Q., and Fung, Y. C. (1989). Relationship between hypertension, hypertrophy and opening angle of
zero-stress state of arteries following aortic constriction. J. Biomech. En gr. 111:325-335.
Masuda, H., Bassiouny, H., Glagov, S., and Zarins, C. K. (1989). Artery wall restructuring in response to
increased flow. Surg. Forum 40:285-286.
Matsumoto, T., and Hayashi, K. (1994). Mechanical and dimensional adaptation of rat aorta to hyperten-
sion. J. Biomech. Engr. 116:278-283.
Matsumoto, T., and Hayashi, K. (1996). Analysis of stress and strain distribution in hypertensive and
normotensive rat aorta considering residual strains. J. Biomech. Engr. 118:62-73.
Matsumoto, T., Itagaki, H., and Hayashi, K. (1994). FEM analysis of stress and deformation in vicinities
of arterial graft anastomosis. Journal ofApplied Biomaterials 5:79-87.
Matsumoto, T., Tsuchida, M., and Sato, M. (1996). Change in intramural strain distribution in rat aorta due
to smooth muscle contraction and relaxtion. Am. J. Phys. 271 :H 1711-H 1716.
Matsumoto, T., Okumura, E., Miura, Y., and Sato, M. (1999). Mechanical and dimensional adaptation on
rabbit carotid artery cultured in vitro. Med. Bioi. Eng. Comput. 37:252-256.
Nerem, B. (1993). Hemodynamics and the vascular endothelium. J. Biomech. Engr. 115:510-514.
Paasche, P. E., Kinley, C. E., Dolan, F. G., Gonaz, E. R., and Marble, A. E. (1973). Consideration of suture
line stresses in the selection of synthetic grafts for implantation. J. Biomech. 6:253-259.
Patel, D. J., and Vaishnav, R.N. (1972). The rheology of large blood vessels. In Berge!, D. H., ed., In
Cardiovascular Fluid Dynamics, volume 2, 1-64. New York: Academic.
Rachev, A., and Hayashi, K. ( 1999). Theoretical study of the effects of vascular smooth muscle contraction
on strain and stress distributions in arteries. Ann. Biomed. Engr. 27:459-468.
Rachev, A., and Kirilova, M. (2001). A Mathematical Model of Arterial Regeneration Over Bioresorbable
Grafts. in preparation.
Rachev, A., Ivanov, T., and Boev, K. (1980). A model for contraction of smooth muscle (translated from
russian). Mech. Camp. Mat. 2:259-263.
Rachev, A., Stergiopulos, N., and Meister, J. ( 1996). Theoretical study of dynamics of arterial wall remod-
eling in response to changes in blood pressure. J. Biomech. 5:635-642.
Rachev, A., Stergiopulos, N., and Meister, J. (1998). A model for geometrical and mechanical adaptation
of arteries to sustained hypertension. J. Biomech. Engr. 120:9-17.
Rachev, A., Manoach, E., and Moore, J. B. J. (2000). A model of stress-induced geometrical remodeling
of vessel segments adjacent to stents and artery/graft anastomoses. J. Theoret. Bioi. 206:429-443.
Rachev, A. (2000). A model of arterial adaptation to alterations in blood flow. J. Elasticity 61(1/3):83-111.
Rembold, C., and Murphy, R. (1990). Latch-bridge model in smooth muscle:[Ca2+] can quantitatively
predict stress. Am. J. Physiol. 259:C251-C259.
Rodbard, S. (1970). Negative feedback mechanisms in the architecture and function of the connective and
cardiovascular tissue. Perspectives in Biology and Medicine 13:507-527.
Rodriguez, E. K., Hoger, A., and McCulloch, A. D. (1994). Stress-dependent finite growth in soft elastic
tissues. J. Biomech. 27:455-467.
Roux, W. (1885). Der ziichtende Kampf der Teile (Theorie der funktionellen Anpassung). In Engelmann,
W., ed., Gesammelte Abhandlungen iiber die Entwicklungsmechanik der Organismen, Volume 1, 137-
422.
Stein, A. A. (2000). Application of continuum mechanics methods to the modeling of growth of biological
tisuues. In Beloussov, L. V., and Stein, A. A., eds., Mechanics of growth and morphologenesis, 148-
173. Moscow: Moscow University press.
Stergiopulos, N. (2000). Using hemodynamics to limit stent restenosis rates: the ares stent principle. In
12th Conference of the European Society ofBiomechanics, 302.
Stewart, S., and Lyman, D. (1988). Finite elasticity modeling the biaxial and uniaxial properties of com-
pliant vascular grafts. J Biomech. Engr. 110:344-348.
Taber, L. A., and Eggers, D. W. (1996). Theoretical study of stress-modulated growth in the aorta. J
Theoret. Biol. 180:343-357.
Taber, L.A., Ng, S., Quesnel, A.M., Whatman, J., and Carmen, C. J. (2001). Investigating Murray's law
in the chick embryo. J Biomech. 34:121-124.
Taber, L. A. (1995). Biomechanical of growth, remodeling, and morphogenesis. Appl. Mech. Rev. 48:489-
545.
Taber, L.A. (1998a). Biomechanical growth laws for muscle tissue. J Theoret. Biol. 193(2):201-213.
Taber, L. A. (1998b). A model for aortic growth based on fluid shear and fiber stresses. J Biomech.
120:348-354.
Thoma, R. (1893). Untersuchungen iiber die Histogenese und Histomechanik des GefojJsystems. Stuttgart:
Enke.
Vaishnav, R.N., and Vossoughi, J. (1983). Estimation of residual strains in aortic segments. In Hall, C. W.,
ed., Biomedical Engineering II: Recent Developments. New York: Pergamon Press. 330-333.
Vaishnav, R.N., Vossough, J., Patel, D. J., Cothran, I. N., Coleman, B. R., and Ison-Franklin, E. L. (1990).
Effect of hypertension on elasticity and geometry of aortic tissue from dogs. J Biomech. Engr. 112:70-
74.
Vorp, D., Raghavan, M., Borovetz, H., Greisler, H., and Webster, M. (1995). Modeling the transmural stress
distribution during healing of bioresorbable vascular prostheses. Ann. Biomed. En gr. 23: 178-188.
Waliszewski, M. W., Stergiopulos, N., and Meister, J. J. (1999). The use of an organ support system for
the study of arterial wall remodeling in vitro. International Journal of Cardiovascular Medicine and
Science 2:93-97.
Zamir, M. (1979). Shear forces and blood vessels radii in the cardiovascular system. Journal of General
Physiology 69:449-461.
Zarins, C. K., Zatina, M.A., Giddens, D., Ku, D. N., and Glagov, S. (1987). Shear stress regulation of
artery lumen diameter in experimental atherogenesis. J Vase. Surg. 5:413-420.
Computational Methods for Soft Tissue Biomechanics
Taras P. Usyk and Andrew D. McCulloch
University of California San Diego, Department of Bioengineering

9500 Gilman Drive La Jolla, CA 92093-0412, USA
E-mail: mcculloc@be-research.ucsd.edu, Home Page: http://cmrg.ucsd.edu/
Abstract. Computational biomechanics provides a framework for modeling the function

of tissues that integrates structurally from cell to organ system and functionally across the
physiological processes that affect tissue mechanics or are regulated by mechanical forces.
We develop an integrative computational strategy for soft tissue based on the finite element
method, using the biomechanics of the heart as a case study.
1 Background
A fundamental goal of physiology is to identify how the cellular and molecular structure of
tissues and organs gives rise to their function in vivo. Correspondingly, a key goal of in sil-
ica physiology is to develop computational models that can predict physiological function from
quantitative measurements of tissue, cellular or molecular structure. Computational modeling
provides a potentially powerful way to integrate structural properties measured in vitro to physi-
ological functions measured in vivo. It also provides a mechanism to integrate biophysical theory
with experimental observation.
In this chapter, we are interested in biomechanical function in general, and cardiac biome-
chanics in particular: i.e. how the cellular and extracellular organization of myocardial tissue is
integrated into the pumping function of the whole heart. For example, how does myocardial fiber
architecture influence the relationship between the biophysics of crossbridge interaction and the
three-dimensional mechanics of the ventricular chambers?
The physics of the heart and other organs are complex. Geometry, structure and boundary
conditions are often irregular and three-dimensional, non-homogenous and time-varying. Con-
stitutive properties and reaction kinetics are typically nonlinear and time-dependent. Beyond
mechanical responses, fundamental physiological functions include electrical, chemical, thermal
and transport processes in cells and tissues. Therefore, computational methods are needed to
model realistically many of these diverse and multidisciplinary processes encountered in biome-
chanics and tissue engineering.
Structurally based models are usually based on in vitro measurements of anatomy, tissue
architecture and material properties, and cell biophysics. Their results must be validated with
measurements from experiments conducted in vivo or in the whole isolated organ. This iteration
between model and experiment also provides the opportunity for numerical hypothesis testing
and in vivo constitutive parameter estimation. Once validated, the computational models have
multidisciplinary applications to problems in medicine, surgery and bioengineering like diag-
nostic imaging, surgical planning and intervention, medical therapy, and biomedicalengineering
design for tissue engineering or medical devices.
274 T.P. Usyk and A.D. McCulloch
In addition to structural integration across scales of tissue organization from muscle and cell
to organ and system, computational models also provide a foundation for functional integration
across interacting biological processes. Computational models have been developed for a variety
of physiological processes that are coupled to biomechanics such as ionic currents and action po-
tential propagation, contractile dynamics, energy metabolism and cell signaling. By developing
a comprehensive model of cardiac biomechanics, we will also have a framework for developing
integrated models of functional interactions such as excitation-contraction coupling, mechano-
electric feedback, mechanoenergetics and mechanotransduction.
2 The Finite Element Method
2.1 Finite Element Modeling
In many problems of bioengineering, solutions to the field equations governing the biophysics
of continua are sought. The geometry is usually complex and three-dimensional with nonho-
mogeneous boundary conditions and anisotropic microstructures. The finite element method is
a popular computational approach to these problems that has been used successfully in diverse
areas, such as:
- Biofluid mechanics, for example, blood flow in arteries;

- Biosolid mechanics, for example, mechanics of bone implants;
- Multiphasic and mixture mechanics, for example, cartilage in joints;
- Bioheat transfer, for example, laser, cryo or radiofrequency ablation;
- Mass transport, for example, facilitated oxygen diffusion;
- Reaction-diffusion systems, for example, action potential propagation;
- Electromagnetic field theory, magneto-encephalography;
- Electrostatics, for example, defibrillation;
- Image modeling, for example, strain analysis from cardiac magnetic resonance imaging;
- Anatomic modeling, for example, torso models for the inverse electrocardiographic problem.
This section reviews the basic principles and approaches of this valuable computational ap-
proach to field problems in three-dimensional continua.
2.2 Discretization of Continuum Problems
Numerical methods for solving field equations must first discretize the solution into a finite num-
ber of unknowns. The commonest approaches to discretization involve subdividing the domain
into collections of a finite number of points, for example, the finite difference method, or dis-
cretizing the solution into a finite number of functions, for example, Fourier analysis or the finite
element method.
In the case of the finite element method, these functions are piecewise, spanning separate
subdomains (the elements). The second feature of the finite element method is that the governing
equations are represented in integral form rather than differential form. Since the integral over the
whole domain is the sum of integrals over each subdomain, the finite element method provides a
mechanism not only for solving the numerical problem but for assembling it.
Computational Methods for Soft Tissue Biomechanics 275
In the general case, an unknown function u satisfies a certain set of partial differential equa-
tions
(1)
over a domain fl, subject to appropriate boundary conditions
(2)
on the boundaries r. The finite element method seeks the approximate solution u in the form
r
u::::: u= LtPiai, (3)
1
where tPi are basis functions prescribed in terms of independent variables (such as spatial coor-
dinates x, y, z ), and some or all of the parameters ai are unknown.
So, we shall seek to cast the equation from which the unknown parameters ai, are to be
obtained in an integral form:
L Fj (u)dfl+ £rj (u)dr = o, (4)
in which F j and fj prescribe known functions or operators.

These integral forms will permit the approximation to be obtained element by element and
an assembly to be achieved by summation:
where neis the domain of each element and re

is its part of the boundary.
Consider, for example, linear partial differential equations and boundary conditions
A(u) =Lu-f=O infl, (6)
B(u) =Mu-p=O onr. (7)

Then, the approximation scheme (4) will yield a set oflinear equations in the form
Ka+f= 0. (8)
For example, the two-dimensional Poisson's equation with zero boundary conditions can be writ-
ten as
cPu a2 u
A(u)=-----f(x,y)=O infl, (9)
ax 2 ay 2
B (u) =u =0 on r. (10)
2.3 Weak Forms
Variational principles. A variational principle specifies a scalar quantity (functional) II,

which is defined by an integral form
II = l F ( u, :xu, ... ) + l dfl E ( u, ! u, .. .) dF, (11)
in which u is the unknown function and F and E are specified operators. The solution to the
continuum problem is the function u, which makes it stationary with respect to small changes
Ju. Thus, for a solution to the continuum problem, the 'variation' is
6II = 0. (12)
The variational principle related to the governing differential equation, can be written as
(13)
If we consider the usual interpolating function expansion

r
u ::::: u = LtPiai = tPa (14)
1
and insert the above into the expression for 6II, we obtain
JII = JaT l tPT A (wa) dfl + JaT fr tPTB (wa) dr = 0. (15)
The above form, being true for all Ja, requires that the integral itself should be zero.
Least squares approximation. Considering the problem of obtaining the stationary point of
II with a set of constraint equations C(u) = 0 in domain fl, we note that the product CTC
must always be a quantity which is positive or zero:
(16)
A functional can thus be written as
(17)
in which a is a penalty factor and we require stationarity for the constrained solution. If the
constraints become simply the governing equations C(u) = A(u), the method becomes the
least squares finite element method.
II= lAT(u)A(u)dfl. (18)

This equation simply requires that the sum of the squares of the residuals of the differential
equations should be a minimum at the solution. It is apparent that we could also obtain the
solution from the equation
(19)
where p is a diagonal matrix in which Pt, P2, ... etc. are positive-valued functions or constants.
This alternative form is sometimes convenient as it allows for varying weight to be assigned to
separate equations and allows additional freedom in the choice of the approximate solution.
Using the linear system of differential equations (6) and (7) and taking a trial function ap-
proximation, we can obtain
II= i [(Ltli) a+ b]T p [(Ltli) a+ b] dfl (20)
and
r5II = i r5aT (Ltli) T p [(Ltli) a+ b] dfl + i [(Ltli) a+ b]T p (Ltli) r5adfl
= 2r5aT [i (Ltli)T p (Ltli) dfla+ i (Ltli)T pbdfl]. (21)
This immediately yields an approximate equation in the usual form
Ka+f= 0. (22)
If we define the energy functional II as
II=~
2
{ (Lu)T A (Lu) dfl +
ln
1n uTpdfl, (23)
the minimization of this functional (which gives the exact solution u u) is equivalent to
1
minimizing
II* = ~ (L (u - u)) T A (L (u - u)) dfl. (24)

2 Q
In the above, L is a self-adjoint linear operator and A and p are prescribed matrices. There-
fore, the approximate solution for Lu approaches the exact one Lu as a weighted least squares
approximation.
Equivalently, if the weight function in a weighted residual method is chosen as Ltli (in the
case of a linear system of differential equations), the weighted residual equations may be written
i
in the form
(Ltli) T p [(Ltli) a+ b] dfl = 0, (25)
or
In (L!P) T p (L!P) df2a+ In (L!P) T pbdf2 = 0. (26)
Weighted residual methods. As the set of differential equations (1) must be zero at each point
of the domain n' it follows that
(27)
where
(28)
is a set of arbitrary weighting functions equal in number to the number of equations (or compo-
nents ofu) involved.
The same integral equation can be written for the boundary conditions
(29)
for any set of functions v. Indeed, the integral statement that
l vTA(u)dfl+ lr vTB(u)dF=O (30)
is satisfied for all v and v is equivalent to the differential equations ( 1) and their boundary
conditions (2).
If the unknown function is approximated by the expansion (3), i.e.
r
u c:::' u = LtPiai = Wa, (31)

1
then it is impossible to satisfy both the differential equation and the boundary conditions in
general. The integral statement allows an approximation to be made if, in place of any function
v, we substitute a finite set of prescribed functions
v=w1, v=wj (j=l. .. n), (32)
where n is the number of unknown parameters ai.

Rewriting (29) we obtain
fnwJA(wa)dfl+ [w/B(wa)dF=O (j=l. .. n). (33)
A ('lt a) represents the residual obtained by substitution of the approximation into the differential
equation and B ( 'lta) the residual of the boundary conditions. The last equation is a weighted
integral of such residuals.
A variety of functions have been proposed for the weights, giving rise to a variety of numer-
ical procedures. The most common choices are as follows.
Point collocation: Wj = 8j where 8j is such that for x "I Xj; y "I yj, Wj = 0 but
fn wjd!? = I (unit matrix). This procedure is equivalent to simply making the residual zero at
n points within the domain and integration is nominal.
Subdomain collocation: Wj = I in nj and zero elsewhere. This makes the integral of the
residual zero over the specified subdomain.
The Bubnov-Galerkin method: w j = IJtj. The basis functions are used as the weighting func-
tions. Then, disregarding the boundary conditions, the Bubnov-Galerkin weighted residual equa-
tion becomes
LIJ!T A(u)dn = o. (34)
Writing a linear system of differential equations as
A (u)::::: Lu +b = 0, (35)
we can rewrite the Bubnov-Galerkin weighted residual equations as
l q,T [(LIJt) a+ b] d!? = o, (L q,TLIJtd!?) a+ l q,Tbd!? = o, (36)
which give rise to the usual matrix form
Ka+f = 0. (37)
For example, the weak form and Galerkin weighted residual equations for two-dimensional
Poisson's equation with zero boundary conditions can be written as
(38)
Note, that w is scalar function, presuming, that the boundary condition is automatically satisfied
by the choice of the function w.
Integrating by parts,
1(--- + ---
n
auaw
ax ax
auaw) dxdy- /,au
ay ay
-wdx- /,au
r ay
-wdy =
r ax n
f(x,y)wdxdy, 1 (39)
l (~~ ~: + ~~ ~;) dxdy = l f(x,y) wdxdy, (40)
where u and w vanish at the boundary. Note that the variable u has disappeared from the integrals
taken along the boundary, and the boundary conditions on that boundary are taken into account
automatically. Such a condition is known as a natural boundary condition.
Allowing an approximation to u in the form
r
u (x, y) = LIJti(x, y) ai, (41)
we can rewrite the integral equation as
(42)
or in matrix form
[K][a] = [f,] (43)
where [K] is called the 'stiffness matrix' and [f] is the 'load (right-hand-side) vector':
K. . =
"1
1(atlfia atlfja
Q X X
atlfi atlfj )
+ ay ay
d d
x y, (44)
(45)
Variational principles for the two-dimensional Poisson's equation with zero boundary condi-
tion can be written as
MI = l [ ~~)
( (a 1~u) ) + ( ~~) (a 1~u) )J dxdy
-l f (x, y) (Ju) dxdy = 0, (46)
HI= (Jai) l [ ~:i)

( (~;) + ( ~:i) (~;)] ajdxdy
-(Jai) l ft/Ji, dxdy = 0 (47)
and in matrix form,

[a]r [K] a= [a]r [f]. (48)
2.4 Steps in the Finite Element Method
The solution of a finite element problem typically involves the following steps.
1. Formulate the weighted residual (weak form).

2. Integrate by parts (or use the Green-Gauss theorem for vector or tensor variables): this re-
duces the derivative order of the differential operator and naturally introduces the derivative
(Neumann) boundary conditions, such as boundary fluxes or tractions.
3. Discretize the problem: discretize the domain into subdomains (elements) and discretize the
dependent variables using finite expansions of piecewise polynomial interpolation functions
(basis functions), weighted by parameters defined at the nodes between adjacent elements.
4. Derive the Bubnov-Galerkin finite element equations: substitute the dependent variable ap-
proximation in the weighted residual integral. Choose the weighting functions to be the in-
terpolation functions- the Bubnov-Galerkin assumption.
5. Compute element stiffness matrices and load vectors: integrate the Galerkin equations over
each element subdomain; integrate the right-hand-side to obtain the element load vectors,
which also include any specified Neumann boundary conditions.
6. Assemble the global stiffness matrix and load vector: add element matrices and load vectors
into a global system of equations. The row and column structure of the global matrix depends
on the node ordering.
7. Apply essential (Dirichlet) boundary conditions: at least one prescribed value of the depen-
dent variables must be specified for a solution; for example, displacements specified on at
least one node. Eliminate the rows and columns corresponding to constrained nodal degrees
of freedom from the global stiffness matrix and transfer the column effects of the prescribed
values to the right-hand-side to achieve the constraint-reduced global stiffness matrix.
8. Solve the global equations for unknown nodal dependent variables using standard algorithms
for Ax= b or A(x)x =b.
9. Evaluate element solutions by interpolating dependent variables and computing their deriva-
tives; for example, stresses or strain energy.
10. Visualize and post-process the solution graphically.
11. Test for convergence: refine the finite element mesh and repeat solution.
2.5 Examples
To illustrate the weighted residual approximation procedure and its relationship to the finite ele-
ment method, we will solve step-by-step a very simple one-dimensional example:
d2 u
A(u) = dx2 = 2, (49)
u(1)=0, u(4)=9. (50)

It is clear, that the solution is simply
u (x) = (x- 1) 2 . (51)
Let us try to obtain the solution of this problem using a weighted residual finite element
method (see Figure 1). The weighted residual equation can be written as
~ 4 ~:~ w -
( 2w) dx = 0. (52)
After integration by parts, we obtain
4 4
/,
( -du
- dw
- - 2w ) dx + [du
-w ] = 0. (53)
1 dxdx dx 1
We define a finite element approximation with four global nodal parameters Ui (i = 1 ... 4)
and three linear elements, as shown in Figure 1. Each element is defined by two nodal parameters
6
u
4 U3 =? 0
1 2 X
3 4
Figure 1. Bubnov-Galerkin finite element method: simple one-dimensional example.
node 1 node3
global nodes: c Ua
element
nodes: U!
•
0
element 1
0
element 2
.
element 3
I
... {
Figure 2. The finite element method: relationship between global nodal parameters U; (i = 1, ... , 4) and
local nodal parameters Uj(j = 1, 2).
Ul and U2. Adjacent elements share global nodal parameters; for example, global parameter u2
is element parameter u 2 of element 1 and u 1 of element 2 (Figure 2). Two linear interpolation
functions for each element are defined in local (element) coordinates as follows: 'Pi (~), i =
1, 2; cp 1 (~) = 1 - ~; cp 2 (~) = ~; ~ E (0; 1). Allow element approximations to u in the form
i = 1, 2. (54)
The Bubnov-Galerkin equation for each element can be written as
!1
2
( -du
-dw
- -2w ) dx+
dxdx
1
2
3
( -du
-dw
- -2w ) dx
dxdx
+ 1 3
4
( du-dw
--
dxdx
- 2 w) dx+ [du
4
-w] =0.
dx 1
(55)
In each element, let

u(x) = uj'<pj' (x) + u~<p~ (x) = ui'Pi (x), i = 1, 2, (56)
w (x) <pi(x), = (57)
where 'Pi (x) are basis functions in global (x) coordinates, defined for each element (e):
<pHx) = 2- x, <p~ (x) = x- 1, x E (1, 2), (58)
<t?i(x)=3-x, <p~(x)=x-2, xE(2,3), (59)
<pUx)=4-x, <p~(x)=x-3, xE(3,4). (60)
The Galerkin equations for each element are
1X2 (-uj'-
x1
d<pe
dx
1 - u~-2
d<pe ) d<p~
dx
- • dx
dx
= 2 1x2
x
<t?idx,
1
(61)
or, in matrix notation,
K, 3
~. - -1X2 dipi
-
d<pj
d d dx, (62)
Xl X X
For example, for element number 1 we obtain

2
1 [ [ 1 -1] [ -1 1 ] (63)
Kii =- }1 -1 1 dx = 1 -1 ,
!l = 2 J 1
2
[2-x] dx= [-(2-xt] = [1].
x- 1 (x- 1) 1 1
2
(64)
Assemble the global stiffness matrix and load vector:
Kij = [Y ~1 H]
0 0 00
+ [El JJ]
0 0 0 0
+ [~ E ~ ] [Y ~2 22 ~ ] .
0 0 1 -1
=
0 0 1 -1
(65)
/;=m+m+m m (66)
Thus, we can obtain a solution from the system of linear equations
(67)
with essential boundary conditions at each end:
U1=u(l)=O, U4=u(4)=9. (68)
That leaves global equations 2 and 3. Solving this constraint-reduced system of two remaining
linear equations, we get
Note, that this is the exact solution at the nodes, though it overestimates the solution between the
nodes.
3 Approximation and Discretization

3.1 Approximating a Field Variable
To approximate a set of points { xk; uk (xk)} by a continuous function, a convenient and popular
method is to use a polynomial expression such as u (x) = a + bx + cx 2 + dx 3 + ... and then
to estimate the monomial coefficients a, b, c, d, ... to obtain a best approximation to the field
variable u(x).
Since high-order polynomials, such as quartics and quintics, tend to oscillate unphysically,
it is helpful to divide a large or complex domain into smaller subdomains and use low-order
piecewise polynomials over each of them- these subdomains are the finite elements (Figure 3).
For example, a field variable u( x) may be represented by several linear elements, as illustrated
in Figure 3.
It is generally necessary to impose constraints to ensure continuity of u across the element
boundaries. Re-parameterizing the linear function from monomial coefficients a and b in the first
element in terms of the nodal values of u at each end of the element (u1 and u2), we write
(70)
where~ E (0, 1) is a normalized measure of distance along the one-dimensional element (Fig-
ure 3).
3.2 Lagrange and Hermite Interpolation

The geometric or dependent field variables of an element may in general be interpolated in two
or three dimensions by
nmax'Ymax
u = L L!li~(e) (6,6,6) · u~(e)' (71)

n=l oy=l
u u = u(x)
u =a+bx u= c+dx u = e+fic
Figure 3. Parameterizing piecewise polynomials in terms of shared nodal parameters automatically ensures
continuity of u across the element boundaries.
where P~(e) (6, 6, 6) are piecewise interpolation functions, u~(e) are nodal parameters at local
node n of element (e). The index "( identifies each of the 'Ymax nodal parameters used to interpo-
late u. For three-dimensional elements, nmax = 8. The interpolation functions P~( e) ( 6, 6, ~3)
may be constructed as tensor products of separate polynomials in each ~k direction. For three-
dimensional rectangular finite elements, these equations may be written as
(72)
0 0 1
Figure 4. The one-dimensional linear Lagrange basis functions.

0 0 0.5
Figure 5. The quadratic Lagrange basis functions.
Linear interpolation. For any variable u, a linear variation between two values may be de-
scribed as
(73)
where the parameter~ is a normalized measure of distance along the curve.
We define
'PI W= 1- ~' cpz (~) = ~' (74)
so that
(75)
where 'PI (~) and cp 2 (~) are the linear Lagrange basis functions associated with the nodal pa-
rameters UI and u 2 (Figure 4).
Quadratic interpolation. A quadratic variation of u over a one-dimensional element requires

three nodal parameters:
(76)
The quadratic Lagrange basis functions are shown in Figure 5. Notice that since 'PI (~) must
be zero at ~ = 0.5 (node 2), 'PI (~) must have a factor (~ - 0.5) and since it is also zero at
~ = 1 (node 3), another factor is(~- 1). Finally, since 'PI(~) is 1 at~ = 0 (node 1), we have
'PI W = 2 (~- 0.5) (~- 1). Similarly, for the other two quadratic Lagrange basis functions,
cpz (~) = -4~ (~- 1), 'P3 (~) = 2~ (~- 0.5).
Cubic Hermite interpolation. All the basis functions mentioned thus far are Lagrange basis
functions and provide C 0 continuity of u across element boundaries but not higher-order conti-
nuity. Often it is desirable to use interpolation that also preserves continuity of the derivative of
u with respect to ~ across element boundaries. A convenient way to achieve this is by including
Figure 6. Cubic Hermite basis functions.
additional nodal parameters, being the derivatives (au I The basis functions are chosen to
a~) n.
ensure that
aui (au)
a~ ~=0 = a~ 1 = u1'
1 aui (au) =
a~ ~=1 = a~ 2
1
Uz'
(77)
and since u is shared between adjacent elements, derivative continuity is ensured. The cubic
Hermite basis functions are thus derived from
(78)
subject to the constraints
u (0) =a= u 1 , u (1) =a+ b + c + d = uz, (79)

aua~ (0)= b = u1, aua~ (0)= b + 2c + 3d =
I
Uz.
1
(80)
Solving these equations, we get
(81)
where the four cubic Hermite basis functions are sketched in Figure 6 and are given by the
equations
'P~ (~) = 1 - 3e + 2e, 'P~ (~) = ~ (~- 1) 2 , (82)

'P~ (~) e
= (3- 2~), 'PH~) = (~- 1). e (83)
1.----------.--------
()
0.5 1
Figure 7. Two-dimensional isoparametric element with linear Lagrange interpolation in one direction and
quadratic Lagrange interpolation in the other.
Two-dimensional elements. Two-dimensional bilinear basis functions are readily constructed

from the products of the above one-dimensional linear functions as
u(~t,6) = LlJin(e) (6,6) · Un( e)' (84)

n=l
where
lJil (6,6) = (/)1 (6)rpl (6)' l[t2 (6 , 6) = (/)2 (6) (/)1 (~2) ' (85)
1Ji3 (6,6) = (/)1 (6)'P2 (6)' l[t4 (6,6) = (/)2 (6) (/)2 (6)' (86)
and the functions 'Pi (~k), (i, k = 1, 2) are defined above by equations (74).
Higher-order, two-dimensional parametric basis functions can be similarly constructed from
products of the appropriate one-dimensional basis functions. For example, a six-noded (see Fig-
ure 7) quadratic-linear element (quadratic in ~1 and linear in 6) would be given by
U (6' 6) = LlJin(e) (6' 6) · Un( e)' (87)

n= l
lJil (6' 6) = rpf (6) rpf (6)' l[t2 (6,6) = rp~ (6) rpf (6)' (88)
l[t3 (6' 6) = rp~ (6) rpf (6)' lft4{6' 6) = rpf (6) rpf (6)' (89)
l[t5 (6' 6) = rp~ (6) rp~ (6)' 1Ji6 (6,6) = rp~ (6) rp~ (6). (90)
A two-dimensional bicubic Hermite element requires four derivatives per node for a total of
16 parameters:
au au a2u
u, (91)
a~l' a6' a6a6'
The need for the second-order cross-derivative term can be explained as follows. If u is cubic
in 6 and cubic in 6 then aula6 is quadratic in 6 and cubic in 6, and aula6 is cubic in 6
and quadratic in 6. Now consider the 1-3 edge as shown in Figure 8. The cubic variation of u
with 6 is specified by the four nodal parameters u1, (au Ia6) 1 , U3 and (au Ia6) 3. But since
aula6 (the normal derivative) is also cubic in 6 along that side and is entirely independent
of these four parameters, four additional parameters are required to specify that cubic. Two of
these are specified by (aula6) 1 and (aula6) 3 , and the remaining two by (a 2 ula6a6) 1 and
(a2 uIa6 a6) 3 . The bicubic interpolation of these nodal parameters is thus given by
4 4
U (6, 6) = L Llji~(e) (6' 6) · U~(e), (92)

n(e)=1r=1
where
1 _
Un(e) - Un(e)'
2
Un(e)
au )
= ( {jC '
3
Un(e)
au )
= ( {jC '
4
Un(e) = (
at<,1a atu<,2 ) n(e) '
2
(93)
<,1 n(e) <,2 n(e)
tJif (6,6) = cp~ (6) cp~ (6)' tli;} (6,6) = cpg (6) cp~ (6)'
w~ (6,6) = cp~ (6) cpg (6), tliJ (6,6) = cpg (6) cpg (6),
tJif (6' 6) = tpt ( 6) cp~ (6) ' tli} (6' 6) = cp~ (6) cp~ (6) '
tli£ (6' 6) = tpt ( 6) cpg (6) ' tJil (6' 6) = cp~ (6) cpg (6) '
tlif (6' 6) = cp~ (6) tpt (6)' tli] (6' 6) = cpg (6) cpt (6)'
tJif (6, 6) = cp~ (6) cp~ (6)' tJil (6, 6) = cpg (6) cp~ (6)'
tJif (6' 6) = tpt (6) tpt (6)' llii (6' 6) = cp~ (6) tpt (6)'
Pi (6, 6) = <p~ (6) <p~ (6), Pi (6, 6) = <p~ (~t) <p~ (6),
and the functions cp{ (~k), (i = 1, 2; j = 0, 1; k = 1, 2, 3) are defined above by equations (82).
Three-dimensional elements. Three-dimensional trilinear Lagrange basis functions are sim-
ilarly constructed from the products of the above one-dimensional linear functions as
8
U (6, 6, 6) = Lljin(e) (6,6, 6) · Un(e)' (94)

n=1
where
lji1 (6' 6, 6) = CfJ1 (6) C{J1 (6) C{J1 (6)' lji2 (6' 6, 6) = CfJ2 (6) CfJ1 (6) CfJ1 (6)'
lji3 (6,6,6) = C{J1 (6) CfJ2 (6) C{J1 (6)' lji4 (6,6,6) = CfJ2 (6) CfJ2 (6) CfJ1 (6)'
lji5 (6' 6, 6) = C{J1 (6) C{J1 (6) CfJ2 (6)' lji6 (6' 6, 6) = CfJ2 (6) CfJ1 (6) CfJ2 (6)'
tli1 (6 , 6, 6) = CfJ1 ( 6) tp2 ( 6) tp2 ( 6) , tlis (6 , 6, 6) = tp2 ( 6) tp2 ( 6) tp2 ( 6) .
The functions 'Pi (~k), (i = 1, 2; k = 1, 2, 3) are defined by equations (74). These eight basis
functions correspond to the eight nodes of a trilinear brick element (Figure 9).
node4
'
node 1 node 2 ~I
Figure 8. Bicubic interpolation.
~~
Figure 9. An 8-noded three-dimensional isoparametric finite element.

A three-dimensional tricubic Hermite element requires eight derivatives per node,
au au au ~u ~u ~u ~u
(95)
u,a6' a6' a6' a6a6' a6a6' a6a6' a6a6a6'
with
8 8
u(6,6,6) = 2: l::.P~(e) (6,6,6) ·u~(e)' (96)

n(e)=1r=1
where
1
Un(e) = Un(e)' un(e)
2
ac
= (au)
<,1 n(e)
3
ac '
' un(e) = (au) <,2 n(e)
(97)
5
un(e) = (
ac
au )
<,3 n(e) ' un(e) = ac ac
6 au ) (
2
<,1 <,3 n(e)

' (98)
a3u )
( a6a6a6
U~(e) = n(e)'
(99)
!Jif (6' 6, 6) = <p~ (6) <p~ (6) <p~ (6)' .Pi (6' 6, 6) = <pg (6) <p~ (6) <p~ (6)'
.Pi (6' 6' 6) = <p~ (6) <pg (6) <p~ (6) ' !Jil (6' 6' 6) = <pg (6) <pg (6) <p~ (6) '
.PJ (6' 6, 6) = <p~ (6) <p~ (6) <pg (6)' .PJ (6' 6, 6) = <pg (6) <p~ (6) <pg (6)'
.Pi (6' 6, 6) = <p~ (6) <pg (6) <pg (6)' .PJ (6' 6, 6) = <pg (6) <pg (6) <pg (6)'
!Jif (6, 6,6) = <p~ (6) <p~ (6) <p~ (6)' .Pi (6, 6, 6) = <p~ (6) <p~ (6) <p~ (6)'
.Pi (6, 6, 6) ='Pi (6) <pg (6) <p~ (6)' .Pi (6, 6, 6) = <p~ (6) <pg (6) <p~ (6)'
!Jil (6' 6, 6) ='Pi (6) <p~ (6) <pg (6)' !Jil (6' 6, 6) = <p~ (6) <p~ (6) <pg (6)'
!Ji'f (6, 6,6) ='Pi (6) <pg (6) <pg (6)' .Pi (6, 6, 6) = <p~ (6) <pg (6) <pg (6)'
!Jil (6,6, 6) = <p~ (6) 'Pi (6) <p~ (6)' lji~ (6,6, 6) = <pg (6) <p~ (6) <p~ (6)'
!Jif (6,6,6) = <p~ (6) <p~ (6) <p~ (6)' !Jif (6,6,6) = <pg (6) <p~ (6) <p~ (6)'
!Jil (6' 6, 6) = <p~ (6) 'Pi (6) <pg (6)' !Jil (6' 6, 6) = <pg (6) 'Pi (6) <pg (6)'
IJif (6' 6, 6) = <p~ (6) <p~ (6) <pg (6)' lji: (6' 6, 6) = <pg (6) <p~ (6) <pg (6)'
.Pt (6, 6,~3) ='Pi (6) 'Pi (6) <p~ (6)' .Pi (6, 6, 6) = <p~ (6) 'Pi (6) <p~ (6)'
.Pi (6' 6' 6) = 'Pi (6) <p~ (6) <p~ (~3) ' .Pt (6' 6' 6) = <p~ (6) <p~ (6) <p~ (6) '
.Pt (6,6, 6) ='Pi (6) 'Pi (6) <pg (6)' .Pt (6,6,6) = <p~ (6) 'Pi (6) <pg (6)'
!Ji:f (6, 6, 6) ='Pi (6) <p~ (6) <pg (6)' .Pi (6, 6, 6) = <p~ (6) <p~ (6) <pg (6)'
tlif (6,6,6) = 'P~ (6) 'P~ (6) 'P~ (6)' cpg (6) 'P~ (6) 'P~ (6)'
t[r~ (6,6,6) =
t[r~ (6' 6, ~3) = 'P~ (6) cpg (6) 'P~ (6)' tlit (6' 6, 6) = cpg (6) cpg (6) 'P~ (6)'
tlif (6 ' 6' 6) = 'P~ (6) 'P~ (6) 'P~ (6) ' tlig (6' 6' 6) = cpg (6) 'P~ (6) 'P~ (6) '
t[r~ (6, 6, 6) = 'P~ (6) cpg (6) 'P~ (6)' t[r~ (6, 6, ~3) = cpg (6) cpg (6) 'P~ (6)'
tlif (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' tJrg (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)'
tlif (6, 6, 6) = 'P~ (6) cpg (6) 'P~ (6)' tlif (6, 6, 6)
= 'P~ (6) cpg (6) 'P~ (6)'
tJrZ (6,6, 6) = 'P~ (6) 'P~ (6) 'P~ (6), tli~ (6,6, 6) = 'P~ (6) 'P~ (6) 'P~ (6),
t[r~ (6' 6' 6) = 'P~ (6) cpg (6) 'P~ (6) ' t[rg (6' 6' 6) = 'P~ (6) cpg (6) 'P~ (6) '
tlil (6' 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' tliJ (6' 6, ~3) = cpg (6) 'P~ (6) 'P~ (6)'
tliJ (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' tliJ (6, 6, 6) = cpg (6) 'P~ (6) 'P~ (6)'
tlil (6,6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' tliJ (6, 6, 6) = cpg (6) 'P~ (6) 'P~ (6)'
tliJ (6, 6,6) = 'P~ (6) 'P~ (6) 'P~ (6)' tlil (6, 6, 6) = cpg (6) 'P~ (6) 'P~ (6)'
tlif (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' t[r~ (6,6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)'
t[r: (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' tlif (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)'
tJr2 (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)' tJrg (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6)'
tliJ (6,6,6) = 'P~ (6) 'P~ (6) 'P~ (6)' t[r~ (6, 6, 6) = 'P~ (6) 'P~ (6) 'P~ (6).
The functions cp{ (~k), (i = 1, 2; j = 0, 1; k = 1, 2, 3) are defined above by equations (82).
3.3 Global-to-Element Mapping
Adjacent elements share global nodal parameters U.1 (see Figure 2), defined at each global node
Ll. Thus, it is necessary to map global nodal parameters U.1, defined at global node Ll, onto local
node n of element e by use of the connectivity matrix Ll(n, e):
Un = U.1(n,e) · (100)
By isa parametric interpolation, it is straightforward to interpolate u( x), simply by defining

the geometric variable x as an interpolation of nodal values itself:
(101)
·'--------- +· •
~=0 ~ _1- - - - - - - - +
~=0
s2
Figure 10. Scaling factors.
It is more accurate to define the global nodal derivative parameters as (aulas)n where sis
arc length in global units, and then to compute the local basis function parameters (au 1a~) n for
each element with respect to its own local coordinates~:
(~~) n = (~~) ~(n,e). (~;) n. (102)
The term (as I aOn is an element scale factor, which scales the arc length derivative of global
node~ to the local derivative of element node n required for the local interpolation (Figure 10).
3.4 Coordinate Systems

In the governing equations, various orthogonal coordinate systems (Figure 11) have been used
to describe the finite deformation of an elastic material in which a point with position vector
R in the undeformed body B moves to a point with position vector r in the deformed body B.
It is sometimes convenient to model the geometry of the region (over which a finite element
solution is sought) using an orthogonal curvilinear coordinate system. Capital letters are used
for coordinates and indices of tensor components referred to B, while lower case letters denote
association with B . There are some particular curvilinear coordinate systems which are useful
for finite element modeling of biological structures, such as blood vessels, cells or the chambers
of the heart (Figures 12 and 13).
e
Rectangular Cartesian Coordinates: { A} = (X' Y, Z) and {(}a} = (X' y' z ) .
Cylindricalpolarcoordinates: {GA} = (R,G,Z) and {(}a}= (r,e,z):
= z.
x = r cos (}, y = r sin (} , z (103)
Spherical polar coordinates: {e A} = (R, e, cf>) and {ea} = (r, 0, ¢) :
x = r cos(} cos¢, y = r sin(} cos ¢, z = r sin¢. (104)
Prolate spheroidal coordinates: { G A} = (A, M, G) and {Oa} = (>.., J.L, (}) :
x = dcosh>..cosJ.L, y = dsinh>..sinJ.Lcose, z = dsinh>..sinJ.LsinO, (105)
where dis focal length of the particular prolate spheroidal coordinate system.
The position vectors R = YReR and r = yrer are defined with respect to the global rectan-
gular Cartesian reference coordinate system. However, it is often more convenient and efficient to
describe the geometry of a curved body using a suitable curvilinear system of world coordinates,
eA. The deformed configuration of the body is given by 0a coordinates.
' \
--
Figure 11. Four coordinate system are used in our finite element method. A rectangular Cartesian
global reference coordinate system {Y1, Y 2 , Y3 } and orthogonal curvilinear coordinate systems
{Eh, 8 2 , 8 3 } are used to describe the geometry. Curvilinear local finite element coordinates
are {~ 1 ,6 , 6}, and locally orthonormal convecting body/fiber coordinates are {X1,X2,X3}
(reproduced from Costa et al. (1996a)).
A) Rectangular Cartesian Coordinates: {E>A}=(X,Y,Z)
v3 ·z~y---~
Y 1 =X
B) Cylindrical Polar Coordinates: {E>A}=(R,E>,Z)
Y3 =Z
Y 1 = R case
Y2 = R sinE>
y3 =Z
C) Spherical Polar Coordinates: {E>A}=(R,E>,<D)
Y 1 = R cos<D case
Y2 = Rcos<D sinE>
Y3 = Rsin<D
Figure 12. Relationship between rectangular Cartesian (A) reference coordinates, Y;, and curvilinear world
coordinates, 8;, for two of the alternative orthogonal coordinate systems used to formulate the
finite element equations: (B) Cylindrical polar; and (C) Spherical polar (reproduced from Costa
eta!. (1996a)).
296 T.P. U syk and A.D. McCulloch
~ = d coshA cosM
'12 = d sinhA sinM cose
't3 = d sinhA sinM sine
Figure 13. Relationship between rectangular Cartesian reference coordinates, Y;, and prolate spheroidal co-
e
ordinates, { A} = {A, M, 8} used here to describe a thick-walled confocal ellipsoidal shell
bounded by inner and outer surfaces of constant A (the dimensionless transmural coordinate) and
truncated at M = 120°. Dimensional scaling is determined by the focus length, d, which has
units of length (reproduced from Costa eta!. (1996a)).
The geometry of the body is discretized, and nodal geometric variables are interpolated us-
ing polynomial functions of normalized finite element coordinates ~k (Figure 11). The fibrous
structure of the anisotropic myocardium is defined using locally orthonormal body/fiber coor-
dinates Xi in which X 1 is aligned with the local muscle fiber axis and lies in the epicardial
tangent (6, 6) coordinate plane. X s lies in the laminar sheet plane and X n is orthogonal to
the sheet plane (Figure 14). The relationship between these coordinate systems is described by a
transformation matrix [M], as shown by Usyk et al. (2000).
3.5 Finite Element Equations
The geometry of an element in B is interpolated as
e= L L [wN] [eN] e; Ju = L L [wN] [ouN] e, (106)

e=l N=n(e)=l e=l N=n(e)=l
where ['lt N] is the matrix of interpolation functions defined as products of independent Lagrange
and Hermite polynomials in each ~k direction for each node, e = (e1, e2, e3) are unit vectors
which describe the curvilinear coordinate system, and [eN] is a matrix-vector of nodal coordi-
nates generalized derivatives at local node N of element (e). A scaling factor [sN] matrix is
defined for each element, so that equations may be written
L L
emax Nmax
e= [wN] [sN] [eLl(N,e)] e, (107)

e=l N=n(e)=l
Xn
Xs
Figure 14. The fibrous structure of the anisotropic myocardium can be defined using locally orthonormal
coordinates defined by the tissue fiber-sheet microstructure and derived in terms of the local
finite element coordinates (modified from Usyk eta!. (2000)).
L L
€max Nmax
8u = [wN] [sN] [8uLl(N,e)J e, (108)

e=l N=n(e)=l
where [BN] = [sN] [BLl (N,e)] and [8uN] = [sN] [8uLl(N,el] .
Using vector notation in the curvilinear coordinate system, we can write
(109)
where
(110)
In components,
I[!N SN 0Ll(N,e)
sk kl l , s=l . .. 3; k,l=l . .. ')'max> (111)
e=l N=n(e)=l
where '/'max is the total number of nodal parameters used to interpolate 8 A, including derivative
terms when Hermite polynomials are used.
(112)
8l]!N l]JN 8H l]JN 8H )

+ 8ts H rkH
Emax Nmax (
'9() = e 0 e ""' ""'. __sk_ _ tk t t SN ()Ll(N,e)
t s~ ~ H act H H acs acr kl l
e=lN=n(e)=l t <, t s <, t r <,
emax Nmax
N SN ()Ll(N,e)
Qtsk
=et®esL L kl l
e=l N=n(e)=l
emax Nmax
=et®esL L [Q([wN])][sNJ[eLY.(N,e)J. (113)

e=l N=n(e)=l
In cylindrical coordinates,
(114)
(115)
8l]!Z/
az
In spherical coordinates,
H1 = Hr = 1, Hz= He= r, H3 = Hq, = rsinB, (116)
(117)
4 Anatomical Models and Kinematics

4.1 Least Squares Fitting for Nodal Geometric Parameters
To apply the general theory of equations ( 18), ( 19), (24) to fit finite element models to anatomical
measurements, we introduce the objective function
F (X) = L 'Yd IIX (~d) - Xdll 2 ' (118)

d=l,D
where Xd is a measured coordinate or field variable, X(~d) is the interpolated value of ~d, which
is defined by the projection of the measured point onto a surface, and 'Y d is the corresponding
weight applied to the data point. The objective function represents the error between the coordi-
nate of a measured anatomical surface point and the corresponding coordinate projected on the
a b
Figure 15. Fitting a one-dimensional mesh to data with (a) two linear Lagrange elements and (b) eight linear
Lagrange elements.
element surface. Weight parameters are all set to one, when all measurements can be assumed to
be equally accurate. As shown above, we can define X as
L L
€max Nmax
X= [wNJ [sN] [xL1(N,e)J e, (119)

e=l N=n(e)=l
or, in components,
L L
€max Nmax
Xs = P:fcSiJX1L1(N,e), s = 1. .. 3; k,l = 1.. ·/max, (120)

e=l N=n(e)=l
L L L
€max Nmax 2
F (X) = /d II [q,N] [sN] [XL1(N,e) Je- xd I (121)

d=l,D e=l N=n(e)=l
A least-squares fit minimizes the objective function:
oF -0 (122)
{)XL1(N,e) - .
J
We can illustrate the fitting algorithm using simple linear Lagrange one dimensional elements.
Figure 15 shows how subdividing two linear elements into eight results in an improved approxi-
mation to artificial data.
4.2 Ventricular Geometry

The mammalian heart consists of four pumping chambers, the left and right atria and ventricles
communicating through the atrioventricular (mitral and tricuspid) valves, which are structurally
connected by chordae tendineae to papillary muscles that extend from the anterior and poste-
rior aspects of the right and left ventricular lumens. The muscular cardiac wall is perfused via
the coronary vessels that originate at the left and right coronary ostia located in the sinuses of
Valsalva immediately distal to the aortic valve leaflets. Surrounding the whole heart is the col-
lagenous parietal pericardium that fuses with the diaphragm and great vessels. These are the
anatomical structures that are most commonly studied in the field of cardiac mechanics. Particu-
lar emphasis here is given to the ventricular walls, which are the most important for the pumping
function of the heart. Most studies of cardiac mechanics have focused on the left ventricle, but
many of the important conclusions apply equally to the right ventricle.
From the perspective of engineering mechanics, the ventricles are three-dimensional thick-
walled pressure vessels with substantial variations in wall thickness and principal curvatures both
regionally and temporally through the cardiac cycle. The ventricular walls in the normal heart
are thickest at the equator and base of the left ventricle and thinnest at the left ventricular apex
and right ventricular free wall. There are also variations in the principal dimensions of the left
ventricle with species, age, phase of the cardiac cycle, and disease.
Ventricular geometry has been studied in most quantitative detail in the dog heart Nielsen et
al. (1991). Geometric models have been very useful in the analysis, especially the use of con-
focal and nonconfocal ellipses of revolution to describe the epicardial and endocardial surfaces
of the left and right ventricular walls. The canine left ventricle is reasonably modeled by a thick
ellipsoid of revolution truncated at the base. The crescentic right ventricle occupies wraps about
180 degrees around the heart wall circumstantially and extends longitudinally about two-thirds
of the distance from the base to the apex. Using a truncated ellipsoidal model, left ventricular
geometry in the dog can be defined by the major and minor radii of two surfaces, the left ventric-
ular endocardium, and a surface defining the free wall epicardium and the septal endocardium of
the right ventricle. Streeter Jr. and Hanna ( 1973) described the position of the basal plane using a
truncation factor fb defined as the ratio between the longitudinal distances from equator-to-base
and equator-to-apex. Hence, the overall longitudinal distance from base to apex is (1 + jb) times
the major radius of the ellipse. Since variations in fb between diastole and systole are relatively
small (0.45 to 0.51 ), they suggested a constant value of 0.5.
The focal length d of an ellipsoid is defined from the major and minor radii (a and b) by
d 2 = a 2 - b2, and varies only slightly in the dog from endocardium to epicardium between end-
diastole (37.3 to 37.9 mm) and end-systole (37.7 to 37.1 mm) (Streeter Jr. and Hanna (1973)).
Hence, within the accuracy that the boundaries of left ventricular wall can be treated as ellipsoids
of revolution, the assumption that the ellipsoids are confocal appears to be a good one. This
has motivated the choice of prolate spheroidal (elliptic-hyperbolic-polar) coordinates (A, Jt, B) as
described earlier (Nielsen et al. (1991) and Young and Axel (1992)). Here, the focal length d
defines a family of coordinate systems that vary from spherical polar when d = 0 to cylindrical
polar in the limit when d -+ oo. A surface of constant transmural coordinate A is an ellipse of
revolution with major radius a = d cosh A and minor radius b = d sinh A. In an ellipsoidal model
with a truncation factor of 0.5, the longitudinal coordinate Jt varies from zero at the apex to 120°
at the base. Integrating the Jacobian in prolate spheroidal coordinates gives the volume of the
wall or cavity:
JJJ((
21r i-'2 >..2
d3 sinh 2 A + sin 2 Jt) sinh A sin Jt) dAdjtdB

0 0 >..1
2Jrd 3 3 >..2
= - 3 -1(1- cOSJt2) cosh A- (1- cos 3 J1,2) coshAI>-. 1 . (123)
Using a truncated ellipsoidal model (Figure 16) as an initial approximation and using the
finite element least squares fitting approach described above, Vetter and McCulloch (1998) built
Figure 16. Initial unfitted prolated spheroidal meshes for the epicardial surface, the left ventricular endo-
cardium, and the right ventricular endocardial free wall and septal walls.
a realistic anatomical model of the geometry of the right and left ventricles of the rabbit heart
(Figure 17). By using prolate spheroidal coordinates to construct surfaces as initial estimates for
the left and right ventricular epicardia and endocardia (Figure 16), Vetter and McCulloch (1998)
reduced the problem to a one-dimensional least squares fit of the .A coordinate alone, which
was approximated on each surface using bicubic Hermite interpolation after the original work of
Nielsen et al. (1991).
Using the same general techniques, it is possible to fit anatomical models to measurements
of other three-dimensional structures such as the atria and blood vessels, etc. Figure 18 shows an
anatomical model of the porcine atria.
4.3 Fiber Architecture

The cardiac ventricles have a complex three-dimensional muscle fiber architecture (for a com-
prehensive review see Streeter Jr. (1979)). Although the myocytes are relatively short, they are
connected such that at any point in the normal heart wall there is a clear predominant fiber axis
that is approximately tangent with the wall (within 3-5° in most regions, except near the apex
and papillary muscle insertions). Each ventricular myocyte is connected via gap junctions at in-
tercalated disks to an average of 11.3 neighbors, 5.3 on the sides and 6.0 at the ends (Saffitz
et al. (1994 )). The classical anatomists dissected discrete bundles of fibrous swirls, though later
investigations showed that the ventricular myocardium could be unwrapped by blunt dissection
into a single continuous muscle 'band' (Torrent-Guasp (1973)). However, more modern histo-
logical techniques showed that in the plane of the wall, the muscle fiber angle makes a smooth
transmural transition from epicardium to endocardium. Similar patterns have been described for
humans, dogs, baboons, macaques, pigs, guinea pigs, and rats. In the human or dog left ventricle,
the muscle fiber angle typically varies continuously from about -60° (i.e. 60° clockwise from
the circumferential axis) at the epicardium to about+ 70° at the endocardium. The rate of change
of fiber angle is usually greatest at the epicardium, so that circumferential (0°) fibers are found
Figure 17. Fitted model of the rabbit heart with epicardial and endocardial surfaces of the left and right
ventricles rendered (from Vetter and McCulloch (1998) ).
Figure 18. Porcine atrial anatomy.

Figure 19. Anatomical model of the rabbit left and right ventricles. 8,351 geometric points and 14,368 fiber
angles were fitted using 36 high-order finite elements (reproduced from Vetter and McCulloch
(1998)).
in the outer half of the wall, and begins to slow approaching the inner third near the trabeculata-
compacta interface. There are also small increases in fiber orientation from end-diastole to systole
(7-19°), with greatest changes at the epicardium and apex (Streeter Jr. et al. (1969)).
A detailed description of the morphogenesis of the muscle fiber system in the developing
heart is not available but there is evidence of an organized myofiber pattern by day 12 in the fetal
mouse heart that is similar to that seen at birth (day 20) (McLean et al. (1989)). Abnormalities
of cardiac muscle fiber patterns have been described in some disease conditions. In hypertrophic
cardiomyopathy, which is often familial, there is substantial myofiber disarray, typically in the
interventricular septum (Maron et al. ( 1987) ).
Regional variations in ventricular myofiber orientations are generally smooth except at the
junction between the right ventricular free wall and septum. A detailed study in the dog that
mapped fiber angles throughout the entire right and left ventricles described the same general
transmural pattern in all regions including the septum and right ventricular free wall, but with
definite regional variations (Nielsen et al. (1991)). Transmural differences in fiber angle were
about 120-140° in the left ventricular free wall, larger in the septum ( 160- 180°) and smaller
in the right ventricular free wall ( 100-120°). A similar study of fiber angle distributions in the
rabbit left and right ventricles has recently been reported (Vetter and McCulloch (1998)). For the
most part, fiber angles in the rabbit heart were very similar to those in the dog, except for on
the anterior wall, where average fiber orientations were 20-30° counterclockwise of those in the
dog.
Using the same least squares methods that they used to fit the ventricular geometry, Vetter and
McCulloch (1998) also fitted a model of fiber architecture into the anatomical model of rabbit
heart. The fitted model was based on about 14,000 histologically measured angles (Figure 19).
Figures 20 and 21 show experimental measurements and model values for fiber angles in different
regions of the left ventricular wall from a dog model (Nielsen et al. ( 1991)) and the rabbit model
(Vetter and McCulloch (1998)). Typical root-mean-squared fitting errors were less than 15-20°.
90
60
Q)
30
....
Q)
0/)
Q)
"'C)
II.), 0
"gh
a.... -30
Q)
..0
t;:::
-60
-90
0% 20% 40% 60% 80% 100%
wall depth%
Figure 20. Fitted fiber angles (lateral LV wall):+ experimental measurements (rabbit);- fitted data (rabbit)
(Vetter and McCulloch (1998);- fitted data (dog) (Nielsen eta!. (1991).
The fibrous architecture of the myocardium has motivated models of myocardial material
symmetry as transversely isotropic. The recognition by LeGrice et al. (1995a) that planes of
cleavage recognized in transverse myocardial sections correspond to parallel, branching laminar
sheets several myocytes thick are the best structural evidence for material orthotropy and have
motivated the development of a models describing the variation of fiber, sheet and sheet-normal
axes throughout the ventricular wall (Legrice et al. (1997)). This also led to the hypothesis that
the laminar architecture of ventricular myocardium is related to the significant transverse shear
strains (Waldman et al. (1985)) and myofiber rearrangement (Spotnitz et al. (1974)) observed
in the intact heart during systole. By measuring three-dimensional distributions of strain across
the wall thickness using biplane radiography of radiopaque markers, LeGrice and colleagues
(LeGrice et al. (1995b )) found that the cleavage planes coincide closely with the planes of max-
imum shearing during ejection, and that the consequent reorientation of the myocytes may con-
tribute 50% or more of normal systolic wall thickening.
4.4 Kinematics
The commonest approach to measuring regional myocardial strains is the use of clinical imaging
techniques, such as contrast ventriculography, high-speed x-ray tomography, magnetic resonance
imaging (MRI) or two-dimensional echocardiography. But the conventional application of these
techniques is not suitable for measuring regional strains because they can not be used to identify
the motion of distinct myocardial points. They only produce a profile or silhouette of a surface,
except in the unusual circumstance when radiopaque markers are implanted in the myocardium
during cardiac surgery or transplantation (Ingels Jr. et al. (1975)). O'Dell and McCulloch (2000)
describe approaches to the non-invasive imaging of regional deformations in the ventricular wall.
The best non-invasive method is magnetic resonance (MTR) tagging, which is now routinely
used to map three-dimensional ventricular strain fields in conscious subjects (Young and Axel
90
+
60 + +
,..- ...,.--
~+ //
30 /
/
/
/
/
/
- ___ ..... /
/
/
-60
-90 +---+----t----t-----t------1
0% 20% 40% 60% 80% 100%
wall depth%
Figure 21. Fitted fiber angles (anterior wall): +local experimental measurements (rabbit);- model fit (rab-
bit) (Vetter and McCulloch (1998);- model fit (dog) (Nielsen eta!. (1991).
(1992)). Figure 22 shows a canine left ventricle model fitted from MR image data using the
least squares finite element method (O'Dell and McCulloch (2000)). The corresponding strain
distributions in the isolated ventricle under passive pressure loading (Figure 23) were calculated
from the deformation gradient tensor, F:
(124)
8xi 8xi a~k

F = Fijei ® ei = axi ei ® ei = a~k axj ei ® ei. (125)
In experimental animals, implantable radiopaque markers are sometimes used for tracking
myocardial motions with high spatial and temporal resolution. While more invasive than MR
methods, the use of implanted material markers has the advantage of higher spatial and tem-
poral resolution and also that chronic 'remodeling' strains can be followed over long periods
as the heart restructures after an insult such as a myocardial infarction (Holmes et al. (1994) ).
Meier et al. (1980a) and Meier et al. (1980b) placed triplets of metal markers 10-15 mm apart
near the epicardium of the canine right ventricle and reconstructed their positions from biplane
cineradiographic recordings. By polar decomposition, they obtained the two principal epicardial
strains, the principal angle, and the local rotation in the region. The use of radiopaque mark-
ers was extended to three dimensions by Waldman et al. ( 1985), who implanted three closely
separated columns of 5-6 metal beads in the ventricular wall. With this technique, it is possible
to find all six components of strain and all three rigid-body rotation angles at sites through the
wall. For details of this method, see the review by Waldman in chapter 7 of Glass et al. ( 1991 ).
An extension of this method for nonhomogeneous strain distributions uses high-order finite el-
ement interpolation of the marker positions to compute continuous transmural distributions of
myocardial deformation, see McCulloch and Omens (1991) (Figure 24).
Figure 22. Original three-dimensional mesh of canine left ventricle and endocardial surface with reference-
state data points.
Figure 23. Regional endocardial circumferential strains during passive Lv filling with respect to the unloaded
reference state.
- "
/L
rlt-d
I
' I
Undefonned state Defonned state
Figure 24. High-order finite elements can be used for nonhomogeneous three-dimensional strain analysis
using the displacements of implanted columns of radiopaque markers.
In order to avoid artifacts in the fitted nonhomogeneous strain fields that may result from
sparse or noisy measurements, the objective function ( 118) can be modified to add smoothing
functions (Young and Axel (1992)). Thus,
F (X) ~ ,'f:v1d IIX (<,) - X, II' + L{,p,(II:~ I ' + v~ II:;, n

+( ~3 1 ~~? 11 + ~4 1 ~~r 11 2 + ~5 118~12~211 2 )} d~,
2
( 126 )
where ~i are weights associated with the smoothing function. The second term penalizes the
components of the difference field u = X - X 0 , for excessive stretch (first derivative term) and
bending (second derivatives terms), where X 0 is the initial estimate of the shape. The smoothing
weights are chosen to regularize the geometric fits without increasing the fitting errors signifi-
cantly beyond the range of the measurement error. The data weights 'Y d can be adjusted to account
for variable confidence in different measured values. As shown above, we can define X as
L L
emax Nmax
X= [wN] [sN] [x,1(N,e)J e (127)

e=l N=n(e)=l
and rewrite the expression for the objective function as
L L L II [wN] [sN] [xL1(N,eJJ e- xdll

emax Nmax 2
F (X)= 'Yd
d=l,D e=l N=n(e)=l
+ ~ J~:~, fn {,,, I a~, (['I'N] [sN] ( [x d(N,•l] ~ [x~(N •l]) e) II'

+1/;2118~2 ([wN] [sN] ([xL1(N,eJJ- [x~(N,eJJ)e)ll2
+1f; 11:;r ([wN] [sN] ([xL1(N,eJJ _ [x~(N,eJJ) e)ll 2
3
+1/;4 11::~ ([wN] [sN] ([xL1(N,e)J _[x~(N,e)]) e)ll 2

+1/;5118~~~6 ([wN] [sN] ([xL1(N,eJJ- [x~(N,el])e)ll2}d~. (128)
5 Three-Dimensional Nonlinear Elasticity

In this section, we derive an efficient comprehensive three-dimensional finite element stress anal-
ysis method for large elastic deformations of nonlinear anisotropic materials. Several features are
included especially for soft tissue biomechanics.
1. The governing equations are formulated in general curvilinear coordinates to model anatomic
structure efficiently and smoothly.
2. The geometric and dependent variables are interpolated using high- and low-order three-
dimensional tensor-product isoparametric basis functions.
3. Muscle fiber and sheet orientations may vary continuously within and between elements for
anisotropic tissues.
5.1 Governing Equations

Consider an elastic body in B containing a region of material with volume V and surface A. The
body is loaded by a surface traction, s, per unit area of the boundary of B, which is in equilibrium
with the internal traction vector, t. The traction vector at the deformed position r in B, measured
per unit area of A in B, may be expressed in terms of the symmetric second Piola-Kirchhoff
stress tensor, P, as t = N · II = N · P · FT, where N is the unit normal to the surface A, and
F is the deformation gradient tensor (Spencer (1980)). The displacement vector u = r - R is
referred to world coordinates Ba in B.
If the body is now subjected to an arbitrarily small displacement 8u, which satisfies compat-
ibility and any displacement boundary conditions specified on A (where bu must be zero), the
principle of virtual work requires that
{ s . 8udA = { t · budA, (129)

jA2 }A
where A 2 is that part of the surface not subject to displacement boundary conditions, so that
}A2
r s . JudA }A2r= N . p . FT . JudA, (130)
or, by means of the Green-Gauss theorem,
r
}~
s. JudA =
hr Div (P. FT. Ju) dV + hr p. FT: [Gradx (JuW dV. (131)
We introduce the transformation matrix M; = a~k I aX I' which describes the relationship
between world coordinates X and element coordinates~ (as described earlier in Section 3.4).
Using this relationship, the deformation gradient tensor can be obtained as
F:::: Gradx (B):::: V'~ (B)· M; Gradx (Ju):::: V'~ (Ju) · M. (132)
Everywhere below V' = V' ~.

The Lagrangian form of Cauchy's first law of motion can be written as
Div (P · FT) = Pof- Pob, (133)
which yields a Lagrangian expression for the virtual work principle,
{ P · (V' (B)· M)T: [V' (Ju) · M]T dV = { Po (b- f)· JudV +{ s · JudA, (134)
h h }~
or
r p. MT. (V' (B))T: MT. [V' (Ju)]T dV = hr Po (b- f). JudV + }~r s. JudA,
h
(135)
where p is the density of the material in B, b is the body force per unit mass, and f is the
acceleration.
To obtain a weak formulation of the equations of motion for problems in finite elasticity,
the stresses, P, are eliminated from equation ( 135) by introducing a hyperelastic strain energy
function, W (E), and using the constitutive relation
P = 2 8W(C)
(136)
ac '
T = (detF)- 1 F · P ·FT. (137)
Using matrix notation and the definition of the vector gradient in curvilinear coordinates (see
equations (107) and (109) above), we can rewrite equation (135) as
I: I=
e=l N=n(e)=l
f]~(e) ~
Ve
[P][M]T ([ QN] [sN] [eL1(N,e)] f et 09 es
: ek 09 e1 [M]T ([QN] [SN] [<5uL1(N,e)]) TdV
~ ~N:~:~, ll~(e) [i. p ([b]- [f]) [WN] [SN] [oud(N,e)] e, e,dV
+ ie [s] [wN] [SN] [<5uL1(N,e)] et · e8 dA], (138)
where
f]n(e) = { 1 when n(e) coincides with global node .1
(139)
L1 0 otherwise
L L
€max Nmax T
f]~(e) [eL1(N,e)]
e=l N=n(e)=l
X~ [P] [M]T [QN] [SN] [sNf [QN]T[M]dV [JuL1(N,e)]
Ve
=I: ~x fl~(e) (i. p([b]- [f]) [wN] [sN] dV [<5uL1(N,e)J
ie
e=l N=n(e)=l e
+ [s] [wN] [SN] dA [<5uL1(N,e) J) . (140)
Finally, we can put this equation into the usual finite element form
L L L L
€max Nmax emax Nmax
f]~(e) [K] [eL1(N,e)] = f]~(e) [q]' (141)

e=l N=n(e)=l e=l N=n(e)=l
f
where
[K] := [P] [M]T [QN] [SN] [sNf [QN]T [M] dV, (142)
Ve
[q] := J p ([b] - [f]) [q,N] [SN] dV + J [s] [q,N] [SN] dA. (143)
Ve Ae
Equations (141) are sufficient to solve for [BL1(N,e)] .

Computational Methods for Soft Tissue Biomechanics 3ll
5.2 Finite Element Solution Method

In the absence of accelerations and body forces (b = f = 0), equations (141) are integrated
using a Gaussian quadrature scheme, and the resulting system of nonlinear elliptic FE equations
are solved for the unknowns (}~,W·e) and p~e) using a Newton iterative method (Oden (1972)).
The non-symmetric element tangent stiffness matrix (Jacobian) may be approximated by forward
differences or analytically and updated at each full Newton iteration. Non-zero contributions to
the constraint-reduced global tangent stiffness matrix are vectorized and solved using a general
linear sparse solver with threshold pivoting. The iterative process is terminated when the sum of
the solution increments and the maximum unconstrained residual are both less than an acceptable
threshold.
5.3 Error Analysis and Solution Convergence
Various a-posteriori error estimators may be used to assess the accuracy and convergence of
finite element approximations. The pointwise difference, eE = E - E , between the strain tensor
obtained numerically, E, and the analytic value of that from a highly converged model, integrated
by Gaussian quadrature throughout the finite element domain,
(144)
to obtain a root-mean-square percentage error for the strain,
(145)
where IIEII is analogous to (144). A root-mean-square error for stress, JIT, may be comp~d
similarly. Numerical solutions can be summarized by the total strain energy of the body, WT,
obtained by integrating the strain energy density over the total finite element domain. The percent
error in total strain energy is computed from
~
II wr = WT - WT lOOw (146)
WT /o,
with WT approximated by the total strain energy for a highly refined finite element mesh in the
absence of exact analytic solutions. The total number of degrees of freedom required for a fully
converged solution (JIWT :S 0.25%) for inflation of a thick-walled prolate spheroid were 340
dof (70 elements, 90 nodes) for linearLagrange interpolation and I 04 dof (3 elements, 8 nodes)
for cubic Hermite interpolation (Costa et al. (1996b )).
5.4 Parallel Finite Element Methods

In simulations involving nonlinear high-order models, evaluation of the element stiffness matri-
ces is the most costly step in terms of time required to obtain the final solution. This additional
computational cost is due primarily to the complex geometry of the elements and the nonlinear
-
s::::
125 t Run time 16 Speedup
....
-EE
·- 100 I
I
;
;
I
I
c.12 ;
;
G) 75 I ::s ;
;
;
~ "C 8 ;
¥
. ... ... ... ..... _____

G)
50
\
~ \ G) ;
;:
\ ;
s::::
~4
;
::s 25
~
0 0
0 12 4 8 16 0 4 8 12 16
Processors Processors
Figure 25. Run times and speedups for a parallel execution of a canine left ventricular model (Costa et
a!. (1996)) using either a modified Newton-Raphson (solid lines) or full Newton (dotted lines)
iterative method for solving the nonlinear equations.
constitutive relation of myocardium. Any element stiffness matrix, however, can be computed
independently of the other element stiffness matrices in the mesh, a fact that may be exploited
using a data-parallel implementation on a scalable parallel processing computer. Parallel imple-
mentation of the finite element procedures was described by Vetter and McCulloch (2000) and
tested using a canine left ventricular model (Costa et al. (1996a)). Performance statistics shown
in Figure 25 illustrate that run times decreased and speedups increased on a distributed memory
parallel computer (Vetter and McCulloch (2000)).
6 Passive Myocardial Mechanics

Two different approaches have been used to develop three-dimensional constitutive models of
resting ventricular myocardium: mechanical testing of myocardial tissue specimens under pre-
scribed homogeneous loading conditions gives stress-strain relationships directly; alternatively,
measurements of regional tissue deformations in isolated or intact whole hearts subject to pre-
scribed loading conditions can be used together with the solution of a boundary value problem
to estimate material constants of an assumed constitutive law in a semi-inverse analysis.
The simplest and commonest test in the former approach is the uniaxial tension test (Mirsky
(1976)). However, such uniaxial stress-strain data are unable to characterize the three-dimensional
constitutive behavior of the myocardium even when independently conducted along different
structural axes (which is very difficult in any case). Biaxial tests measure force and displacement
(stress and strain) along orthogonal fiber and cross fiber axes, and the results of such tests in my-
ocardium were reported first by Derner and Yin (1983). Constitutive parameters estimated from
biaxial experiments suggest that canine ventricular myocardium is 1.5-3.0 times stiffer in the
fiber direction than in the cross-fiber direction and that this ratio varies with load and location
across the thickness of the wall (Novak et al. (1994)).
It remains uncertain how biaxial properties measured in isolated tissue slices are related to
the properties of the intact ventricular wall. The irregularity of isolated myocardial tissue and
variation of fiber direction through the specimen complicate in vitro measurements. Moreover,
shearing deformations are an important component of three-dimensional mechanics in the intact

heart (Waldman et al. (1985)), but the usual biaxial protocols do not test responses to shearing
with respect to fiber and crossfiber axes. A recent study by Dokos et al. (2000) describes a
novel shear testing device that is capable of applying simultaneous shear loads in two orthogonal
directions to small isolated specimens of ventricular myocardium while the resultant forces are
measured.
Guccione et al. ( 1991) used a different approach to identifying constitutive parameters. They
sought a myocardial material law that was representative of the fully three-dimensional loading
conditions of the intact heart. They prescribed the deformations of a cylindrical model to repro-
duce axial, circumferential and torsional strains measured in isolated arrested dog hearts (McCul-
loch et al. (1989)). They then used a nonlinear least-squares optimization algorithm to identify
the parameters of a transversely isotropic exponential strain energy function that minimized the
errors between computed boundary tractions and the known pressure loading conditions on the
experimental preparation. The resulting model was able to reproduce three-dimensional transmu-
ral strain distributions subsequently measured with radiopaque markers (Omens et al. (1991) ),
even though the original data used in the optimization were only two-dimensional strains at one
point on the epicardial surface. Alternatively, one can minimize errors between strains estimated
from the computational model subjected to physiologic loading pressures and the corresponding
measured strains. The finite element method is particularly well suited for this inverse analy-
sis due to its ability to incorporate the three-dimensional geometry, fibrous tissue architecture,
pressure boundary conditions, and nonlinear material properties of the tissue. If it can be shown
that a model accurately describes experimentally measured three-dimensional strains, then this
provides confidence in the estimated stresses, which are important for understanding myocardial
growth and remodeling in physiological and pathophysiological conditions. Alternatively, iden-
tifying the source of discrepancies between estimated and measured strains by carefully testing
the assumptions of the model can provide a valuable means of gaining insight into the essential
features which determine three-dimensional cardiac mechanical function.
Resting myocardium is frequently modeled as a finite elastic material using a hyperelastic
pseudo-strain energy function (Fung (1981)). The components of the second Piola Kirchhoff
stress, Pij, are related to the Lagrangian Green's strain Eij, through the pseudo-strain energy,
(ow + ow)
W:
1 -1
(147)
Pij = 2 oEij oEji - pCij '
where Cij is the right Cauchy-Green deformation tensor (C = 2E + 1). For the special case of
incompressibility p is the hydrostatic pressure, which does not contribute to the deformation of
an incompressible material and must be computed from the governing equations of motion and
boundary conditions.
6.1 Isotropic Constitutive Equations
The simple isotropic Mooney-Rivlin strain-energy potential is a linear function of the first two
principal invariants of the right Cauchy-Green deformation tensor:
p
w= c1 (I1 - 3) + c2 (h - 3) - 2 (Is - 1). (148)
It has been widely used for large deformation analysis of nonlinear rubber elasticity. In the con-
strained formulation, the material is assumed to be incompressible, thus introducing the Lagrange
multiplier p, which must be included as another dependent variable in the analysis in order to
maintain the kinematic constraint that the volume ratio I 3 (the third principal invariant of C)
must remain one.
If C 2 = 0, the Mooney-Rivlin formulation reduces to the well-known neo-Hookean law,
derived for rubber from statistical mechanics by Treloar. If C 2 > 0, C 1 should be > 5C2 to
avoid non-convex behavior.
The Blatz-Ko model is a simple one-parameter law for compressible rubber-like behavior:
(149)
The Mooney-Rivlin and Blatz-Ko laws are primarily useful for testing and validation since some
useful analytic solutions exist (Hill (1973) and Costa et al. (1996a)). They are fairly poor ap-
proximations for most soft tissues (Schmid et al. ( 1995)) but can reproduce the responses of
some synthetic biomaterials such as polyurethane implants (Simpson et al. (2001)).
For soft tissues, including myocardium, Demiray (1976) was among the first to use the
isotropic exponential strain-energy function of the first principal invariant of the right Cauchy-
Green deformation tensor, originally proposed by Blatz in 1969:
p
W = C (exp (b (I1 - 3))- 1)- 2 (h- 1). (150)
For many soft tissues, an exponent of approximately five to ten has been found to work well.
For passive canine myocardium, Hunter et al. (1988) used a value of 5.1.
The isotropic power law is a popular special case of a more general form suggested by Ogden
( 1984 ), in which the strain energy is a function of the principal stretch ratios,
w = ~ (>-r + >-~ + >-~) - ~ (h - 1), (151)
where !3 = >.1>.2>.3 = 1.
There is no requirement for the powers a to be integers, and values around two are typical for
rubber-like materials. However, there is also no particular advantage to non-integer values at the
high magnitudes required for most soft tissues. For example, Bogen et al. ( 1980) used powers of
12-18 for myocardium.
6.2 Slightly Compressible Materials
Strain energy functions for myocardial tissue have traditionally been of the incompressible type.
However, the incompressibility constraint can cause numerical difficulties, and, moreover, bulk
myocardium itself is actually slightly compressible under physiological loads (Yin et al. (1996)),
presumably a reflection of the redistribution of vascular, lymphatic and possibly interstitial fluid
during compression. Peng and Chang (1997) showed that the introduction of the penalty func-
tion is equivalent to the modification of the strain energy function from an incompressible type
to a nearly incompressible one. The strain energy function for nearly incompressible, slightly
compressible materials is usually written as two separate terms: a shear term W 8 and a bulk term
Wb,
(152)
Most of the incompressible strain energy functions proposed previously (without the Lagrange
multiplier term) can be used as W 8 • For the bulk term, the general form
fa
wb =- Jp(13)d13, (153)
1
was proposed, where p (13) is the hydrostatic pressure as a function of the volume ratio 13 • The
pressure can be derived from a penalty function, such as
1
-p= -g(h), (154)
e
where e is the penalty parameter and g (h) is the penalty function. Two forms are widely used.
The first one is logarithmic function
g(h) = lnh (155)
The other form is simply

g (h)= h -1. (156)
Using the logarithmic penalty function, it is possible to obtain nearly incompressible behavior
with a penalty factor representative of the empirically observed bulk modulus. The Mooney-
Rivlin constitutive equation (148), for example, may be rewritten as
W = Ct (It- 3) + C2 (!2- 3)- (Ct + 2C2) lnh + Ccompr (h lnh- h + 1). (157)
In the slightly compressible formulation, values for the bulk modulus, Ccompr, of the order
of 100 kPa produce nearly isochoric behavior at physiological stresses for typical soft biomate-
rials. Note that additional terms involving ln h are present to ensure that stresses are zero in the
unstrained state.
Equations (150) and (151) can be rewritten in the forms
W = C (exp (b (It- 3))- blnh- 1) + Ccompr (h ln13- h + 1), (158)
W = !!:. (,W +A~+ A~- a ln 13) + Ccompr (h ln h - h + 1). (159)

a
6.3 Transversely Isotropic Laws

Humphrey et al. ( 1990a) and Humphrey et al. ( 1990b) determined the functional form of a trans-
versely isotropic polynomial strain energy function directly from biaxial tests in which the first
and fourth strain invariants were independently varied:
w = C1 (It - 3) + C2 (It - 3) 2 + C3 (14 - 1) 2 + C4 (14 - 1) 3

+ c5 (It- 3) (!4- 1)- ~ (h- 1). (160)
For transverse isotropy, the fourth strain invariant 14 is simply the square of the stretch ratio
in the direction of the preferred 'fiber' axis, about which rotations do not change the elastic
response. Novak et al. (1994) identified the five material parameters for this law from tests of
canine myocardium isolated from three layers of the left ventricular free wall: subepicardium,
mid-myocardium and subendocardium.
Humphrey and Yin ( 1987) proposed a three-dimensional form for the strain-energy potential
as the sum of an isotropic exponential function of the first strain invariant and a second expo-
nential function of the fiber stretch ratio (equivalently of the fourth invariant for transversely
isotropic symmetry):
p
W = C1 (exp (b1 (h- 3))- 1) + Ct (exp (bt (14- 3))- 1)- 2 (h- 1). (161)
They identified parameters from biaxial tests in isolated canine left ventricular myocardium. The
isotropic part of this expression has also been used to model the myocardium of the embryonic
chick heart during the ventricular looping stages, with coefficients of 0.02 kPa during diastole
and 0.78 kPa at end-systole, and exponent parameters of 1.1 and 0.85, respectively (Lin and
Taber (1994)).
Gupta et al. (1994) used powers of 22-32 in a constrained orthotropic model fitted to biaxial
tests of scarred and spared (remote) sheep myocardium two weeks after myocardial infarction.
In their model, the stretch ratios were resolved with respect to circumferential, longitudinal and
transmural (radial) axes, respectively.
We have often used exponential strain energy functions, which are variations of the formula-
tions proposed by Fung ( 1981) for skin, arteries and other soft tissues. The transversely isotropic
and orthotropic forms are functions of squared Lagrangian strains resolved with respect to local
muscle fiber (f), crossfiber (c) and sheet (s) axes as defined by the laminar fiber architecture
of the myocardium (Legrice et al. (1997)). These orthogonal axes reduce to fiber-crossfiber-
transmural axes when the sheet angle is zero, and reduce further to circumferential-longitudinal-
transmural axes when the fiber angle is also zero:
1 p
W = 2c(exp(Q) -1)- 2 (h -1), (162)
Q = bffEJt + bxx (E~c + E;s + E~s + E;c) + btx (EJc + E~f + EJs + E;f). (163)
Parameters for the transversely isotropic form have been estimated for passive canine (Guccione
et al. (1991)), rabbit (Vetter and McCulloch (2000)), rat (Omens et al. (1993)) and murine (Weis
et al. (2001)) myocardium. In each case, the coefficient C simply scales the stresses.
Humphrey's constitutive equation (160) in a slightly compressible formulation can be rewrit-
ten as
w= c1 (h - 3) + c2 (h - 3) 2 + c3 (14 - 1) 2 + c4 (14- 1) 3
+C5 (11 - 3) (14 - 1)- C1ln h + Ccompr (h In 13- h + 1). (164)
Equations (161) and (162) can be rewritten in the forms
w= cl (exp(bl (Jl- 3))- bllnh -1)

+Ct (exp (bt (14- 3))- bt lnh -1) + Ccompr (h lnh- 13 + 1), (165)
1
W = 2C (exp (Q) -1) + Ccompr (h lnh- h + 1). (166)
Again, in the slightly compressible formulation, values for the bulk modulus, Ccompr. of the order
of 100 kPa produce nearly isochoric behavior at physiological stresses for typical soft tissues.
6.4 Orthotropy
Most finite element models to date have used transversely isotropic material laws for myocardium.
As described by Legrice et al. ( 1997), ventricular myofibers are organized into branching laminae,
whose orientation varies through the wall and whose presence suggests an alternate interpreta-
tion. Based on this assumption, regional variations in myocardial cross-fiber stiffness must be
interpreted as inhomogeneity of material properties. To extend the constitutive model to material
orthotropy we begin by defining a system of locally orthogonal base vectors (Cartesian) defining
a fiber-sheet-normal coordinate system {X f, X 8 , Xc} with one axis parallel to the muscle fibers,
one parallel to the sheet and perpendicular to the fibers and the third normal to the sheet plane
(Figure 26). Then, the exponential strain energy function used in (162) can be generalized to
1 p
W = 2c (exp (Q)- 1)- 2 (h- 1), (167)
or, for the slightly compressible case,
1
W = 2C (exp (Q) -1) + Ccompr (h lnh- Is+ 1), (168)
where
Q = bf!EJ! + bccE;c + bssE;s + bcs (E;s + E;c}

+bts (EJs +E;f) +bfc (EJc +E;f) · (169)
Parameters for this orthotropic form have been estimated for intact canine ventricular my-
ocardium at end-diastole (Usyk et al. (2000)). The six separate exponents in the orthotropic form
adjust the elastic response to strains resolved with respect to local fiber-crossfiber-sheet coor-
dinates (f, c, s). Usyk et al. (2000) suggested that, in consideration of the myocardial laminar
structure, the two interlaminar shear moduli, btc and bcs• should be equal. Methods to measure
these properties have recently been developed by Dokos et al. (2000).
Finite element model analysis (Usyk et al. (2000)) showed that orthotropic resting material
properties with decreased interlaminar tensile and shearing stiffnesses did improve agreement
with measured diastolic strains in the beating canine heart.
6.5 Boundary Value Problems
Three-dimensional finite elements have been used to obtain solutions to a variety of quasi-static
equilibrium problems in finite elasticity assuming negligible body forces (Costa et al. (1996a)).
These problems were chosen to test different aspects of the method relevant to cardiac mechanics
and for the existence of analytic solutions for FE model validation.
Xn
Figure 26. Myocardial laminar fiber-sheet architecture.

80
60
ca
11.
..11::
-
rn
~ 40
( /)
>-
-'=
tJ
::I
~ 20 "(<!2
o+---~~~==~=-~--~~~
1.0 1.1 1.2 1.3
Stretch Ratio
Figure 27. Equibiaxial extension of a thin rectangular sheet. Analytic fiber stress and cross-fiber stress versus
stretch ratio for an isotropic neo-Hookean material (broken line) and a transversely isotropic
exponential material (solid lines) indicating fiber stiffness greater than crossfiber stiffness. Finite
element solutions (symbols) are shown at intermediate load steps (from Costa eta!. (1996)).
The stress-extension relationship (Figure 27) for equibiaxial stretch of a 1.0 x 1.0 x 0.1
em rectangle, a common experimental protocol for mechanical tissue testing, demonstrates the
expected greater stiffness in the fiber direction than in the cross-fiber direction for the transversely
isotropic material, whereas the neo-Hookean material was isotropic and less nonlinear. One finite
element was used, with trilinear interpolation of rectangular Cartesian coordinates and uniform
p (25 degrees of freedom) and 8-point Gaussian quadrature integration.
A low-order FE cylinder model (radii 1.0 and 1.49 em) was subjected to axial extension
(10.7%) and torsion ( -2.6°/cm), computed to approximate passive inflation (1.7 kPa) of a trans-
versely isotropic equatorial section of the canine left ventricle. Fiber angle varied linearly across
the wall from 75° on the endocardium to -45° on the epicardium. Numerical solutions were
accurate throughout the element as represented by the non-monotonic fiber stress and the trans-
murally decreasing crossfiber stress (Figure 28, triangles).
When torsional constraints were removed from the finite element model, the endocardial sur-
face twisted relative to the epicardium, giving rise to a longitudinal gradient in circumferential-
radial transverse shear strain. A transmural gradient arose in the negative circumferential-longi-
tudinal shear strain ( -0.11 on the endocardium to 0.01 on the epicardium), which had been
small and uniform ( -0.04). Endocardial crossfiber stress increased and fiber stress became more
nonuniform (Figure 28, circles).
Guccione et al. (1991), Costa et al. (l996a) and Costa et al. (1996b) constructed an ax-
isymmetric model of the passive canine left ventricle incorporating a realistic longitudinal ge-
ometric profile and fiber distribution and transversely isotropic material properties (151). The
model showed good agreement with transmural distributions of three-dimensional strains mea-
sured in the isolated dog heart except for an insufficient gradient in longitudinal strain and an
over-estimation of the magnitude of circumferential-radial shear strain. More recently, Vetter
10
o+-~--.-~--.-~--.-~--.-~--.
0 20 40 60 80 100
endocardium epicardium
Undeformed Radial Position (%)
Figure 28. Transmural distribution of fiber stress and crossfiber stress for inflation of an anisotropic cylinder
model of passive myocardium. Numerical solutions were obtained using one low-order cylindri-
cal finite element (symbols represent Gauss point solutions at the midlength cross-section). When
torsion was constrained so that radial segments remained radial, numerical stresses (triangles)
matched the analytic solutions (solid lines). When torsion constraints were released, inflation
caused the endocardium to twist relative to the epicardium, increasing the nonuniforrnity of the
stresses (dashed lines indicate numerical solutions with circles at Gauss point locations). From
Costa et al. (1996).
and McCulloch (2000) described a model of passive filling in the rabbit heart based on a very de-
tailed anatomic reconstruction of right and left ventricular geometry and fiber architecture. The
model reproduced measured epicardial strains in this species well.
7 Active Contractile Mechanics
7.1 Ventricular Hemodynamics
The most basic mechanical parameters of the cardiac pump are blood pressure and volume flow
rate, especially in the major pumping chambers, the ventricles. From the point of view of wall
mechanics, the ventricular pressure is the most important boundary condition. Ventricular filling
immediately following mitral valve opening (MVO) is initially rapid because the ventricle pro-
duces a diastolic suction as the relaxing myocardium recoils elastically from its compressed sys-
tolic configuration below the resting chamber volume. The later slow phase of ventricular filling
(diastasis) is followed finally by atrial contraction. The deceleration of the inflowing blood re-
verses the pressure gradient across the valve leaflets and causes them to close (MVC). Valve clo-
sure may not, however, be completely passive, because the atrial side of the mitral valve leaflets,
which unlike the pulmonic and aortic valves are cardiac in embryological origin, have muscle
and nerve cells and are electrically coupled to atrial conduction (Sonnenblick et al. (1967)).
Ventricular contraction is initiated by excitation, which is almost synchronous (the duration
of the QRS complex of the ECG is only about 60 msec in the normal adult) and begins about
0.1 to 0.2 sec after atrial depolarization. Pressure rises rapidly during the isovolumic contraction
phase (about 50 msec in adult humans), and the aortic valve opens (AVO) when the developed
pressure exceeds the aortic pressure (afterload). Most of the cardiac output is ejected within the
first quarter of the ejection phase before the pressure has peaked. The aortic valve closes (AVC)
20-30 msec after AVO when the ventricular pressure falls below the aortic pressure owing to the
deceleration of the ejecting blood. The dichrotic notch, a characteristic feature of the aortic pres-
sure waveform and a useful marker of aortic valve closure, is caused by pulse wave reflections in
the aorta. Since the pulmonary artery pressure against which the right ventricle pumps is much
lower than the aortic pressure, the pulmonic valve opens before and closes after the aortic valve.
The ventricular pressure falls during isovolumic relaxation, and the cycle continues. The rate of
pressure decay from the value P0 at the time of the peak rate of pressure fall until mitral valve
opening is commonly characterized by a single exponential time constant, i.e.
(170)
where P 00 is the (negative) baseline pressure to which the ventricle would eventually relax if
MVO were prevented (Yellin et al. (1986)). In dogs and humans, T is normally about 40 msec,
but it is increased by various factors including elevated afterload, asynchronous contraction asso-
ciated with abnormal activation sequence or regional dysfunction, and slowed cytosolic calcium
reuptake to the sarcoplasmic reticulum associated with cardiac hypertrophy and failure. The pres-
sure and volume curves for the right ventricle look essentially the same; however, the right ven-
tricular and pulmonary artery pressures are only about a fifth of the corresponding pressures
on the left side of the heart. The intraventricular septum separates the right and left ventricles
and can transmit forces from one to the other. An increase in right ventricular volume may in-
crease the left ventricular pressure by deformation of the septum. This direct interaction is most
significant during filling (Janicki and Weber ( 1980) ).
The phases of the cardiac cycle are customarily divided into systole and diastole. The end of
diastole-the start of systole-is generally defined as the time of mitral valve closure. Mechan-
ical end-systole is usually defined as the end of ejection, but Brutsaert and colleagues proposed
extending systole until the onset of diastasis (see the review by Brutsaert and Sys (1989)) since
there remains considerable myofilament interaction and active tension during relaxation. The dis-
tinction is important from the point of view of cardiac muscle mechanics: the myocardium is still
active for much of diastole and may never be fully relaxed at sufficiently high heart rates (over
150 beats per minute). We retain the traditional definition of diastole, but consider the ventricular
myocardium to be 'passive' or 'resting' only in the final slow-filling stage of diastole.
7.2 Ventricular Pressure-Volume Relations

During the last 25 years, the ventricular pressure-volume relationship (Figure 29) has been ex-
plored extensively, particularly by Sagawa, Suga and colleagues, who wrote a comprehensive
book on the approach (Sagawa (1988)). The isovolumic phases of the cardiac cycle can be rec-
ognized as the vertical segments of the loop, the lower limb represents ventricular filling, and the
upper segment is the ejection phase. The difference on the horizontal axis between the vertical
isovolumic segments is the stroke volume, which expressed as a fraction of the end-diastolic vol-
ume is the ejection fraction. The effects of altered loading on the ventricular pressure-volume re-
lation have been studied in many preparations, but the best controlled experiments have used the
End-
20 systole Ejection
-ro
0..
~
16
.2 c: () c:
'-"' 12 E .Q
·- 0
E:=
Q) :::l- :::l ()
...... - ct! -ro
::J 0 >< 0 ....
en > ct! > -
en o- 0 c:
Q)
......
8 .!!!.~ -I)) 0
()
0..
Filling
0
0 5C 100 150 200
Volume (ml)
Figure 29. The ventricular pressure-volume loop.
isolated cross-circulated canine heart in which the ventricle fills and ejects against a computer-
controlled volume servo-pump.
Changes in the filling pressure of the ventricle (preload) move the end-diastolic point along
the unique end-diastolic pressure-volume relation (EDPVR), which represents the passive fill-
ing mechanics of the chamber that are determined primarily by the thick-walled geometry and
nonlinear elasticity of the resting ventricular wall. Alternatively, if the afterload seen by the left
ventricle is increased, stroke volume decreases in a predictable manner. The locus of end-ejection
points (AVC) forms the end-systolic pressure-volume relation (ESPVR), which is approximately
linear in a variety of conditions and also largely independent of the ventricular load history.
Hence, the ESPVR is almost the same for isovolumic beats as for ejecting beats, although con-
sistent effects of ejection history have been well characterized (Hunter et al. (1988)). Connecting
pressure-volume points at corresponding times in the cardiac cycle also results in a relatively
linear relationship throughout systole with the intercept on the volume axis lfo remaining nearly
constant. This leads to the valuable approximation that the ventricular volume lf(t) at any in-
stance during systole is simply proportional to the instantaneous pressure P(t) through a time-
varying elastance E( t):
P (t) = E (t) (lf (t) -Yo). (I 71)
The maximum elastance Emax, the slope of the ESPVR, has acquired considerable signifi-
cance as an index of cardiac contractility that is independent of ventricular loading conditions. As
the inotropic state of the myocardium increases, for example, with catecholamine infusion, Emax
increases, and with a negative inotropic effect such as a reduction in coronary artery pressure, it
decreases (Figure 30).
20
-
co
16
-
C1..
~
Q)
12
.....
:::J
(/)
(/)
Q) 8
.....
a..
4
50 100 150 200

LV Volume (ml)
Figure 30. The time-varying elastance approximation of ventricular pumping function.
The area of the ventricular pressure-volume loop is the external work (EW) performed by the
myocardium on the ejecting blood:
I
ESV
EW = P(t)dV (172)
EDV
Plotting this stroke work against a suitable measure of preload gives a ventricular function
curve, which illustrates the single most important intrinsic mechanical property of the heart pump
(Figure 31 ). In 1914, Patterson and Starling ( 1914) performed detailed experiments on the ca-
nine heart-lung preparation, and Starling summarized their results with his famous 'Law of the
Heart', which states that the work output of the heart increases with ventricular filling. The so-
called Frank-Starling mechanism is now well recognized to be an intrinsic mechanical property
of cardiac muscle.
The primary determinants of the end-diastolic pressure-volume relation (EDPVR) are the
material properties of resting myocardium, the chamber dimensions and wall thickness, and
the boundary conditions at the epicardium, endocardium and valve annulus (Gilbert and Glantz
(1989)). The EDPVR has been approximated by an exponential function of volume (Gaasch and
LeWinter (1994)), though a cubic polynomial also works well. Therefore, the passive chamber
stiffness dP/ dV is approximately proportional to the filling pressure. Important influences on
the EDPVR include the extent of relaxation, ventricular interaction and pericardia! constraints,
and coronary vascular engorgement. The material properties and boundary conditions in the sep-
tum are important since they determine how the septum deforms (Glantz et al. (1978) and Glantz
and Parmley (1978)). Through septal interaction, the end-diastolic pressure-volume relationship
of the left ventricle may be directly affected by changes in the hemodynamic loading conditions
of the right ventricle. The ventricles also interact indirectly since the output of the right ventricle
is returned as the input to the left ventricle via the pulmonary circulation. Slinker and Glantz
increased contractility (e.g.

adrenergic agonist)
...··························"""
......·
// . · · · .· · · · · · · · · ·: ~;~:·~=·: :ractility
/ ...········
/ (e.g. heart failure)
"Preload" (EDV or EDP)

Figure 31. Starling's Law of the heart (the Frank-Starling mechanism).
(1986), using pulmonary artery and venae caval occlusions to produce direct (immediate) and
indirect (delayed) interaction transients, concluded that the direct interaction is about half as sig-
nificant as the indirect coupling. The pericardium provides a low friction mechanical enclosure
for the beating heart that constrains ventricular overextension (Mirsky and Rankin ( 1979)). Since
the pericardium has stiffer elastic properties than the ventricles (Lee et al. ( 1987) ), it contributes
to direct ventricular interactions. The pericardium also augments the mechanical coupling be-
tween the atria and ventricles (Maruyama et al. (1982)). Increasing coronary perfusion pressure
has been seen to increase the slope of the diastolic pressure-volume relation (an 'erectile' effect),
Salisbury et al. (1960) and May-Newman et al. (1994).
7.3 Active Contraction
Cardiac muscle mechanics testing is far more difficult than skeletal muscle testing mainly owing
to the lack of ideal test specimens like the long single fiber preparations that have been so valu-
able for studying the mechanisms of skeletal muscle mechanics. Moreover, under physiological
conditions, cardiac muscle cannot be stimulated to produce sustained tetanic contractions due
to the absolute refractory period of the myocyte cell membrane. Cardiac muscle also exhibits a
mechanical property analogous to the relative refractory period of excitation. After a single iso-
metric contraction, some recovery time is required before another contraction of equal amplitude
can be activated (Figure 32).
Unlike skeletal muscle, in which maximal active force generation occurs at a sarcomere
length that optimizes myofilament overlap c~ 2.1 JLm), the isometric twitch tension developed
by isolated cardiac muscle continues to rise with increased sarcomere length in the physiolog-
ical range (1.6-2.4 JLm). Early evidence for a descending limb of the cardiac muscle isometric
length-tension curve was found to be caused by shortening in the central region of the isolated
muscle at the expense of stretching at the damaged ends where the specimen was tethered to the
Sarcomere
length, ~-tm Sarcomere isometric
2.1
2.0
1.9
Tension,
mN
2.0
1.0
100 200 300 400 500 600 700
Figure 32. Isometric Testing.
test apparatus. If muscle length is controlled so that sarcomere length in the undamaged part of
the muscle is indeed constant, or if the developed tension is plotted against the instantaneous
sarcomere length rather than the muscle length, the descending limb is eliminated (ter Keurs et
al. (1980)). Thus, the increase with chamber volume of end-systolic pressure and stroke work is
reflected in isolated muscle as a monotonic increase in peak isometric tension with sarcomere
length. The increase in slope of the ESPVR associated with increased contractility is mirrored by
the effects of increased calcium concentration in the length-tension relation. The duration as well
as the tension developed in the active cardiac twitch also increases substantially with sarcomere
length.
The relationship between cytosolic calcium concentration and isometric muscle tension has
mostly been investigated in muscle preparations in which the sarcolemma has been chemically
permeabilized. Because there is evidence that this chemical 'skinning' alters the calcium sensi-
tivity of myofilament interaction, recent studies have also investigated myofilament calcium sen-
sitivity in intact muscles tetanized by high frequency stimulation in the presence of a compound
such as ryanodine that open calcium release sites in the sarcoplasmic reticulum. Intracellular
calcium concentration was estimated using calcium-sensitive optical indicators such as fura. The
myofilaments are activated in a graded manner by micro Molar concentrations of calcium, which
binds to troponin-C according to a sigmoidal relation (Ruegg (1988)). Half-maximal tension in
cardiac muscle is developed at intracellular calcium concentrations of w- 6 to w- 5 M (the Ca50 )
depending on factors such as species and temperature (Bers (1991)). Hence, relative isometric
tension T0 /Tmax may be modeled (Tozeren (1985) and Hunter et al. (1988)) using
To
(173)
Tmax [cat + [Ca]~0 ·
The Hill coefficient (n) governs the steepness of the sigmoidal curve. A wide variety of values
have been reported but most have been in the range 3 to 6 (Kentish et al. (1986) and Backx et
A B
_:q .e
()
g 5 ~ 5
Q)
> >
0 60 120 0 60 120
Force(%) Force(%)
Figure33. Isotonic testing: (A) the results of an isovelocity release experiment conducted during a twitch;
(B) cardiac muscle force-velocity relation corrected for viscous forces of passive cardiac muscle
which reduce shortening velocity.
al. ( 1995) ). The steepness of the isometric length-tension relation compared with that of skeletal
muscle is due to length-dependent calcium sensitivity. That is, the Ca 50 (M) and perhaps n too,
both change with sarcomere length, L.
The isotonic force-velocity relation of cardiac muscle is similar to that of skeletal muscle,
and A.V. Hill's well-known hyperbolic relation is a good approximation except at larger forces
greater than about 85% of the isometric value (Figure 33a):
v 1- TjT0
(174)
Vmax 1 + cT/To ·
The maximal (unloaded) velocity of shortening is essentially independent of preload, but

does change with time during the cardiac twitch and is affected by factors that affect contrac-
tile ATPase activity and hence crossbridge cycling rates. de Tombe and ter Keurs (1992) using
sarcomere length-controlled isovelocity release experiments found that viscous forces impose a
significant internal load opposing sarcomere shortening. If the isotonic shortening response is
adjusted for the confounding effects of passive viscoelasticity, the underlying cross-bridge force-
velocity relation is found to be essentially linear (Figure 33b).
Cardiac muscle contraction also exhibits other significant length-history-dependent proper-
ties. An important example is 'deactivation' associated with length transients. The isometric
twitch tension redeveloped following a brief length transient that dissociates cross bridges reaches
the original isometric value when the transient is imposed early in the twitch before the peak
tension is reached. But following transients applied at times after the peak twitch tension has
occurred, the fraction of tension redeveloped declines progressively since the activator calcium
has fallen to levels below that necessary for all crossbridges to reattach (ter Keurs et al. (1980)).
A number of models of active tension development in cardiac muscle have been proposed. In
essence they may be grouped into three categories: (i) time-varying elastance models include the
essential dependence of cardiac active force development on muscle length and time (Arts et al.
(1979), Chadwick (1982) and Taber (1991)); (ii) 'Hill' models, in which the active fiber stress
development is modified by shortening or lengthening according to the force-velocity relation,
so that fiber tension is reduced by increased shortening velocity (Arts et al. ( 1982) and Nevo and
Lanir (1989)); (iii) Fully history-dependent models are usually based on A.F. Huxley's cross-
bridge theory (Panerai (1980), Landesberg and Sideman (1994) and Landesberg et al. (1996))
which yields a system of partial differential equations as functions of time and cross bridge po-
sition. Many of the early models were based on skeletal muscle models of Hill (Hill ( 1938) and
Hill (1970)). However, Hill's model considers tetanic contraction only, and hence is inappropri-
ate for describing cardiac muscle mechanics (Fung (1981)). According to Panerai (1980), Wong
(1971) and Wong (1972) was the first to employ the sliding filament theory to model the mechan-
ics of cardiac muscle. Wong generalized Huxley's model (Huxley (1957)) of the skeletal muscle
cross-bridge to partial and length-dependent activation. Panerai used Huxley's original model
and incorporated length-dependent activation in a first order kinetic equation describing Ca2+-
troponin C interaction. Instead of considering individual myofilaments, Tozeren ( 1985) proposed
a 'continuum' model of cardiac muscle contraction. Tozeren generalized Hill's equation to par-
tial activation to describe active fiber tension as a function of fiber strain, strain rate and time
after onset of contraction. In these studies, model predictions were validated by experimental
length-tension or force-velocity relations during contractions in which overall muscle length was
controlled. Panerai accounted for the appreciable internal shortening that occurs during isomeric
contractions at the expense of lengthening in the damaged muscle at the clamped ends (Kruger
and Pollack (1975) and ter Keurs et al. (1980)).
Continuum models typically compute the active tension developed by a cardiac muscle fiber
from the peak intracellular Ca2+ ion concentration, the time after onset of contraction and the
sarcomere length history (Guccione and McCulloch (1993)). Model contraction is driven by a
free calcium transient that is independent of length. The number of actin sites available to react
with myosin is determined from the total number of actin sites (available and inhibited), free
calcium and the length-history-dependent association and dissociation rates of two Ca2+ ions
and troponin concentration.
Hunter et al. (1998) described intracellular concentration according to
Cai (t) = Cao + (Camax- Ca 0 ) _t_ (1- _t_) ,

Tea
exp
Tea
(175)
where Ca 0 is the resting concentration of intracellular concentration.

To account for the Ca2+ -troponin C interaction in these models of cardiac muscle mechanics
and energetics, Panerai derived the rate equation
(176)
from classical chemical kinetics to describe the binding of two calcium ions to independent sites
on the troponin molecule, where Ac represents the concentration of actin that is free to react with
myosin, Aco is a constant reflecting the total amount of actin present in the muscle, and c 1 and
c2 are the association and dissociation rate constants, respectively.
Hunter et al. (1998) introduces a non-dimensional parameter z (0 ::::; z ::::; 1) to model
tropomyosin kinetics, representing the proportion of actin sites available for crossbridge bind-
ing. Tropomyosin movement resulting from TnC-Ca 2 + binding controls the availability of these
sites and, since tension increases exponentially with a first-order rate constant that depends on
the level of calcium, they proposed a first-order model for z in the form
~: = ao [ Cb~~~J n (1- z)- z] , (177)
where a 0 is the rate constant of tropomyosin movement and [Ca]5 0 and n are Hill parameters
fitted to the equilibrium relationship between z and [Ca2+] at a given sarcomere length. A sig-
moidal response curve (Hill-type equation) is used to describe the steady-state developed tension
as a function of calcium concentration [Ca2+],
(178)
Zss = [Cat + [Ca]~0 '
where Zss takes a value between 0 and 1, [Ca]5 0 is the calcium concentration required to produce
50% of the peak contraction and n is the Hill parameter describing the shape of the sigmoidal
curve. Note that equation (178) is also known as a 'Hill' equation, not to be confused with
Hill's equation for the force-velocity relation (174). Sarcomere length dependence is included
by describing developed maximum tension as a function of A., the extension ratio of sarcomeres
Clo/lref, where lref is resting sarcomere length):
To= Tref [1 + (30 (A.- 1)]. (179)
From experimental observations in rats, Tref = 125 kPa (the reference tension when A. = 1)
(Hunter et al. (1998)). The term {3 0 describes myofilament cooperativity:
1 dT0
f3o = T:-
ref
d \ = 1.45.
/1
(180)
For less then full activation, the dependence on calcium under isometric conditions may be
approximated by
To= Tref [1 + (3 0 (A.- 1)] Zss· ( 181)
The length-dependence of the parameters n and [Ca]5 0 is approximated in a similar form:
n = nrer[1 + fJ1 (A.- 1)], (182)
pC5o = pC50ref [1 + fJ2 (A.- 1)], [Ca] 50 = 10 6 -pCso in f1M. (183)
In order to fit the experimental data that Kentish et al. (1986) obtained from skinned rat right
ventricular muscle, the parameters of the model were chosen as follows: nref = 4.25; f3o = 1.45;
fJ1 = 1.95; fJ2 = 0.31; pC5oref = 5.33; Tref = 125kPa; Camax = 4.3 f1M, the maximal
intracellular calcium concentration.
7.4 Ventricular Activation
Myocardial systolic contraction is usually modeled at the continuum scale by defining the Cauchy
stress tensor in world coordinates as the sum of the passive three-dimensional stress tensor T (P l
derived from the strain energy function (see previous section) and an active stress tensor T(a)
following Hill's original assumption:
(184)
The components Ti~) of the active stress tensor in curvilinear world coordinates were derived
from a diagonal stress tensor T active referred to local fiber-sheet coordinates (X 1, X 8 , Xn) using
a rotation matrix q which define the relation between the curvilinear world coordinate system and
the local fiber-sheet coordinate system using deformed fiber and sheet angles (Usyk et al. (2000)):
(185)
The components of active tensor T active are a function of peak intracellular calcium [Ca] i and
sarcomere length. The parameters of the active model are based on experimental measurements
of sarcomere length and tension (Guccione and McCulloch (1993) and Hunter et al. (1998)).
Windkessel models for aortic and pulmonary arterial impedance may be coupled to ventricu-
lar pressures and volumes to compute hemodynamic boundary conditions. Ventricular volume
constraints are imposed during the isovolumic phases when all the valves are closed.
Previous active models of myocardial tissue have used the assumption that activation is uni-
axial (along the fibers) and active stress matrix T active has only one nonzero component Tf'jtive,
which is a function of peak intracellular calcium [Ca] i and sarcomere length. In more recent stud-
ies, transverse active stresses r;;tive and T~~tive have been added as a function of axial stress
Tf'jtive and local transverse (.As, An) and axial >. 1 strains (Usyk et al. (2000)) to be consistent
with biaxial experimental tests in barium-contracted rabbit myocardium (Lin and Yin ( 1998) ).
To reproduce these biaxial observations we made the following assumptions: (i) part of the
significant transverse active stress measured in isolated tissue is associated with the observed 12-
150 angular dispersion (a) of local myofiber axes about the regional mean (Karlon ( 1998)); (ii)
part of the transverse active stress is generated by unknown mechanisms acting at the myofibril
level, probably associated with cross-bridge geometry (Schoenberg (1980)), and this transverse
myofibril stress development is a constant fraction (k) of the axial tension. The experimentally
measured active biaxial stresses were also affected by the measured 18° 'splay' in mean myofiber
angles through the thickness of the left ventricular experimental tissue slices used in biaxial tests
oftonically activated ventricular muscle specimens (Lin and Yin ( 1998) ).
Fiber, sheet and sheet-normal stresses acting along the mean myofiber, sheet and sheet-
normal axes in a ensemble of cells is obtained by integrating myofibril tractions over the dis-
tribution of myofiber angles f( B):
Tactive =II
e 11
[Q · Tlocal · QT] · f (B) <p (!1) df.tdB. (186)
This equation corresponds to our first assumption about angular dispersion of local fibers about
the regional mean. The active stress tensor T 1ocai in local fiber coordinates for equation (186)
consists of only fiber component T}jcai, which is a function of peak intracellular calcium [Ca]i
and sarcomere length. The parameters B and 11 are angles that describe the relationship between
the local myofiber axis and the mean fiber axis; f( B) is the fiber orientation probability density
distribution which can be described using a von Mises distribution; rp(f.t) also is a density distri-
bution which can be described as rp(f.L) = 1/ (27r), 11 E [1, 21r]; Q defines the relation between
the mean fiber-sheet coordinate system and local cell coordinate system:
Xfiber-sheet = QXIocai. (187)

Iff (B) is jO (B) in the undeformed reference state, then
(188)
where Ai are fiber, sheet and normal-sheet extensions. In general, the function fO (B) may also
depend on the polar angle J.l about the mean fiber axis. It has been assumed that the local myofiber
orientation may be approximated by a von Mises distribution (Karlon (1998)),
(189)
where ko is a concentration parameter, and l 0 is the modified Bessel function of order zero, given
by
J
2rr
Io (k) = 2~ exp [ko cos¢] d¢. (190)

0
The concentration parameter ko can be related to the angular dispersion by
1 A (k ) = h (ko) (191)
1 o Io (ko)'
where a is the angular dispersion and h is the modified Bessel function of order one. Series
expansions for the modified Bessel functions are given by
L
00
Io (k 0 ) = (r!) - 2 (0.5k 0 ) 2 r, (192)

r=O
= :2:: [(r + p)!r!]- 1 (0.5k

00
Ip (ko) 0 )
2 r+p, p = 1, 2, ... (193)
r=O
The concentration parameter k 0 may assume any value between 0 and oo with smaller values
associated with more dispersed distributions and larger values associated with narrower distribu-
tions.
We may rewrite equation ( 186) as
Using our second assumption about transverse force development mechanisms at the myofib-
ril level, associated with cross-bridge geometry, we obtain r;~cal and T~':-,cal as a constant fraction
( k) of the active fiber stress T}jcal. In this case, the active stress tensor in local fiber coordinates
can be written as
Tlocal = (195)
A. 250
0
N" 200
E
..!:!
s 150
"'"'
!:
II)
100
]i 50
0
1-
0
1.00 1.05 1.10 1.15 1.20
B. 250
N" 200
E
~ 150
"'"'
!]i 100
50
0
1-
0
c. 250
1.00 1.05 1.10 1.15 1.20
i' 200
u
]! 150
~
"'"'!!! 100
u;
]i 50
0
1-
0 +--~e.;t!t~~;:g:::;:~
1.00 1.05 1.10 1.15 1.20
Stretch Ratio
Figure 34. Total stress in fiber (squares) and cross-fiber (circles) directions for equibiaxial (A), off-biaxial
(B), and uniaxial (C) stretch protocols from Lin and Yin (1998) compared with model results
(solid lines).
Tlocal = kTlocal As Tlocal = kTlocal An

ss !! At' nn ff Af ·
(196)
The factors As I A1 and An I A1 are included to reflect the assumption that k is determined by the
angle between the crossbridge and the thin filament which changes as a function of transverse
l
lattice spacing (Schoenberg (1980)).
After integration of equation (194), we can obtain the active stress tensor in mean fiber-sheet
coordinates:
TJ'jtive 0. 0
Tactive = [ 0 r;sctJve 0 (197)
0 0 ractive
nn
Usyk et al. (2000) computed total stressesfor simulated biaxial testing with a three-dimensional
finite element model that included the splay of mean fiber angles through the thickness of the test
specimen and passive material parameters consistent with resting biaxial stress-strain tests in the
rabbit (Lin and Yin (1998)). The comparison in this study shows excellent agreement with ex-
perimental fiber and transverse active stresses (Lin and Yin (1998)) for uniaxial, equibiaxial and
off-biaxial stretch protocols using constant values of k = 0.3 and IJ" = 12° (Figure 34).
02,-------- 0.2 02
• experiment
0.1 - - - biaxial stress 01 01
- uniaxial stress
00 0.0 0.0
"LU"
I
~
LU"' LU~
-0.1 -0.1
-0.1
-0.2 -0.2 -0.2
-0.3 -03---- -03

-02 0.0 02 04 06 0.8 10 1.2 -02 00 0.2 0.4 06 0.8 1.0 12 -02 00 0.2 0.4 06 0.8 1.0 12
Wall depth% Wall depth% Wall depth%
0.3~--- · - - ---! 0.2 , - - - - - - - - - - - - - - - ,
0.2
0.1 -
00- ~
. - l
-0.1
-0.2
-02
1------------
0.0 0.2 0.4 0.6 0.8 1.0 1.2 -02 0.0 0.2 04 0.6 08 10 12 -0.2 0.0 0.2 0.4 06 08 1.0 12
Wall depth% Wall depth% Wall depth%
Figure35. Systolic strains in the canine left ventricle for uniaxial (solid lines) and biaxial (dotted lines)
models of active contraction (reproduced from Usyk et al. (2000)).
In Figure 35, the components of the strain tensor in fiber-sheet coordinates with orthotropic
passive properties and both uniaxial and biaxial active properties are represented. The agreement
with the experimental data (Usyk eta!. (2000)) was significantly better for the model with biax-
ial active stress than uniaxial active stress. Contour maps of systolic fiber and crossfiber strain
components (Figure 36) in a model of the canine left ventricle are also more realistic for the
transversely isotropic model.
During isovolumic contraction and relaxation, the model required volumetric constraint equa-
tions to calculate the pressure boundary condition at each time step, to keep cavity volume of the
left and right ventricles constant. These equations for each time step can be written as
(198)
where Ck is a penalty factor that was adjusted to produce good numerical convergence without
permitting significant volume errors. The ventricles were coupled to two-element Windkessel
models in order to simulate pulmonary and arterial input impedances so that ventricular pressures
could be calculated during ejection (Figure 37). The equations for this model can be written in
terms of pressure p and volume V as
cdp !!_ _ dV
(199)
dt + R- dt.
The parameters C and R were chosen for the pulmonary and aortic circulations to obtain
physiological systolic pressures. These parameters agreed qualitatively with experimentally es-
timated parameters and were different for the left and right ventricles, as expected. Figure 38
shows the pressure-volume loops from the model for the left and right ventricles.
Transversely Isotropic Uniaxial
c
~
...Cll
(/)
.c
u:::
Figure 36. Systolic fiber and crossfiber strains in a model of the canine left ventricle assuming uniaxial
and biaxial models of active contraction. Note the unphysiological crossfiber stretching on the
epicardium when active stress development is assumed to be one-dimensional.
Pressure-Volume Relationship (Left Ventricle)
10 15 20 30 35
Volume (an- J
Figure 37. Left ventricular pressure-volume loop from the model.

Pressure-Volume Relallonsl'up (Left Ventricle) Pressure-volume relationship (nght ventncle)
20 10 .-------------~
,..----- 8 8
I
-16
c <ii' c
~
"" ~ 6
;
Cll
10
4
0.. ~ 2
D A
o"-______A
0
10 IS 20 25 30 3S 10 15 20 26 30 35
Volume(an ) Voh.me (em~)
Figure 38. Pressure-volume loops from the computational model of the canine left and right ventricles.
8 Pathological Cases
8.1 Acute Myocardial Ischemia
Impaired systolic function in the normally perfused myocardium adjacent to an ischemic region,
the functional border zone, is thought to result from mechanical interactions across the perfusion
boundary. Adetailed computational model of the canine left ventricle has been used to investigate
how segment orientation and vessel involved affect regional strains in the functional border zone
and whether altered stresses associated with a step transition in contractility can explain the
functional border zone (Mazhari et al. (2000)).
The three-dimensional finite element model of left ventricular mechanics was based on com-
prehensive measurements of canine geometry and myofiber architecture (Nielsen et al. ( 1991))
and used to simulate filling and ejection using measured diastolic and systolic left ventricu-
lar pressures (Figure 39). To model the ischemic myocardium, myofiber Ca2+ sensitivity and
calcium-activated maximal tension development were reduced abruptly across the perfusion
boundary in the ischemic region with a step transition across the perfusion boundary, Allen and
Orchard (1987) (Figure 40). The shape of the ischemic region and location of the perfusion
boundary were based on experimental measurements (Mazhari et al. (2000)).
Figure 41 shows fiber and crossfiber end systolic strain components for normal and ischemic
(left anterior descending and left circumflex coronary artery occlusions) states of the canine
left ventricle (Mazhari et al. (2000)). The model was able to explain experimentally observed
strain distributions in the functional border zone with no need to postulate a zone of intermediate
contractile properties.
8.2 Myofiber Disarray: Hypertrophic Cardiomyopathy
Localized disarray of the myofibers is a characteristic of various cardiac pathologies, notably

hypertrophic cardiomyopathy, which also includes ventricular wall thickening and regional ab-
normalities in systolic function, especially on the septal wall. The MLC2v-ras transgenic mouse
expresses human oncogenic ras (a cell growth signaling molecule) in the ventricular myocytes
Figure 39. Anterior view of a three-dimensional model of the canine left ventricle. Epicardial and endo-
cardial fiber orientation vectors are shown. The ischemic region corresponding to left anterior
descending coronary artery occlusion is the hatched area.
1.6 1 .8 2 2 .2
Sarcomere Length (pm)
150
(i125
0. /
..:.:: /
";;100
-
I
...
Ill
Q>
75 I
(/) I
Q> 50
> I - Baseline
; I
u 25
c:( - -I scherria
0
0 5 10 15 20 25
[Cct1i (J.ll\'l
Figure 40. Active systolic tension as a function of intracellular calcium concentration and sarcomere length
for normal and acutely ischemic (2 mins occlusion) myocardium.
CONTROL LAD OCCLUSION LCx OCCLUSION
.....
c
·~
en
Q)
.Q
LL
.....
c
·~
en
Q)
.Q
~
til
u
e
til
-0.10 0.00 0.10
Figure 41. Mid wall fiber and cross fiber strain distributions from the three-dimensional model during base-
line, circumflex (LCx) and left anterior descending (LAD) coronary artery occlusions (modified
from Mazhari et al. (2000)).
and displays a phenotype characteristic of hypertrophic cardiomyopathy with septal hypertrophy

and focal myocyte disarray.
Random foci of disarray were modeled in the septal wall by finite elements with modified sys-
tolic material properties. The pattern of disarray was reconstructed from statistical distributions
of disarray obtained by Karlon et al. (2000), who measured local fiber orientation and angular
deviations in 21 x 21 grids from each of 9 transmural tissue sections spaced equally through the
septal wall thickness in MLC2vlras transgenic mice (Figure 42). Using these original data, Usyk
et al. (2001) calculated the number of disarrayed three-dimensional regions and the distribution
of their volumes in seven mice. Based on these statistics, they built a stochastic region-growing
algorithm to generate random three-dimensional regions of focal disarray in the model, occupy-
ing about 25% (100) of the finite elements in the septal wall of the model. Each finite element
in the septal wall of the model was assumed to be either entirely normal or entirely disarrayed
(piecewise constant variation).
The influence of two different mechanisms associated with microstructural alterations of dis-
arrayed tissue were examined with the model:
(i) a local increase in myofiber angular deviation;

(ii) a local decrease in sarcomere length (which was also observed experimentally).
Each of these structural alterations was modeled individually by altering material properties of
disarrayed elements in the finite element mesh.
Mechanism 1: Altered generation of crossjiber tension. Dispersion of fiber orientation was

measured by Karlon et a!. (2000) as the local angular deviation (angular equivalent of standard
Figure 42. Measurements of local fiber orientation and angular dispersion in tissue sections from wildtype
and ras transgenic mouse hearts showing focal regions of myofiber disarray (Karlon (1998)).
ro
ro 200
a..
a.. - Normal
X 80
-"' - - SL vi - - Normal
vi 160 --AD
Ill
-
Ill ~ 60 - - SL
~ 120 - - SL+AD / iii --AD
__..// /
iii Q; 40 - · - SL+AD
Q; ...... / / / .a ~/
./ /
.a 80 ,... / <.=
/
_......_..../ ?/
<.=
-
--
.2 40 /
/
~ ......
Ill
Ill
0
20 .;::::
0 .-I _... ... .,.,.
t; -#
iii :::.-.. 0
>- 0 .2
(/)
B
000 0 05 010 0 15 0 20 0 25 Ill
>.
Fiber strain (/) 0.00 0.05 0.10 0.15 0.20 0.25
Cross fiber strain
Figure 43. Systolic stress-strain curves for normal and disarrayed tissue subject to biaxial deformations. The
influence of reduced sarcomere length (SL), increased angular dispersion (AD) and both of these
factors combined (SL+AD) are shown (reproduced from Usyk et al. (2001)).
deviation) in myofiber orientation. To determine the net contribution of a family of myofibers

with a known distribution of orientations, a superposition of local stress contributions was per-
formed. Using the equations in Section 7.4 with measured average angular deviations of 12° for
normal myocardium and 25° for disarrayed tissue (Figure 42), one obtains a ratio of systolic
crossfiber/fiber stress of 0.45 in normal muscle versus 0. 70 in disarrayed elements.
Mechanism 2: Reduction in sarcomere length. Development of systolic tension in cardiac

myocytes is dependent upon activator calcium concentration and sarcomere length (ter Keurs et
al. (1980)). The isometric length-tension curve describes the tension development at different
sarcomere lengths for given concentrations of calcium. Measurements indicated that sarcom-
ere lengths in areas of myofiber disarray were reduced both in systole and diastole (Karlon et
al. ( 1998) ). This suggests that these myocytes operate at a lower point of the sarcomere length-
tension curve, thus generating less tension and shortening. Systolic isometric fiber stress develop-
ment in the model was computed as a function of activator calcium concentration and sarcomere
length according to a model of length-dependent activation (Guccione and McCulloch (1993)).
The reference state sarcomere length of all normal elements was set to 2.0 I J.tm (Hunter et al.
(1998)). Areas of disarray were modeled by reducing reference state sarcomere lengths to 1.65
J.lm according to the experimental observations.
To illustrate how both of these mechanisms influence myocardial mechanics, simple cuboidal
blocks of tissue were modeled with normal material properties and with disarray material prop-
erties. The resulting stress-strain relationships are shown in Figure 43, and illustrate significant
differences in the systolic mechanical properties of control and disarrayed myocardium.
When these properties were used in a finite element model of the mouse heart with random
regions of disarray (Usyk et al. (2001)), average systolic strain in disarrayed regions was sig-
nificantly reduced on the septal wall, consistent with experimental observations (Karlon et al.
(2000)). The subendocardial systolic strains associated with disarrayed tissue found near the LV
side of the septal wall were reduced less. There were also significantly smaller average torsional
shear strains associated with areas of disarrayed tissue near the right ventricular side of the septal
surface, again consistent with observation. The effects of reduced sarcomere length and reduced
anisotropy separately were about the same in magnitude and sign, suggesting that both mecha-
nisms contribute approximately equally to observed septal wall dysfunction in this animal model
of hypertrophic cardiomyopathy.
Summary. In summary, we have seen that the finite element method, when appropriately
developed and customized, can provide a framework for computational soft tissue biomechanics
that can be used to integrate both between theory and experiment, and structurally from cell
to tissue to organ and organ system. This integration can also be extended further to include
functional coupling between the many physiological processes that interact with mechanics, such
as cell growth and signaling, metabolism, transport and electrophysiology.
References
Allen, D. G., and Orchard, C. H. (1987). Myocardial contractile function during ischemia and hypoxia.
Circ. Res. 60(2): 153-68.
Arts, T., Reneman, R. S., and Veenstra, P. C. (1979). A model of the mechanics of the left ventricle. Ann.
Biomed. Engr. 7:299.
Arts, T., Veenstra, P. C., and Reneman, R. S. (1982). Epicardial deformation and left ventricular wall
mechanics during ejection in the dog. Am. J. Physiol. 243:H379.
Backx, P. H., Gao, W. D., Azan-Backx, M.D., and Marban, E. ( 1995). The relationship between contractile
force and intracellular [Ca2+] in intact rat cardiac trabeculae. J. Gen. Physiol. 105(1):1-19.
Bers, D. M. (1991). Excitation-Contraction Coupling and Cardiac Contractile Force. Dordrecht: Kluwer.
Bogen, D. K., Rabinowitz, S. A., Needleman, A., McMahon, T. A., and Abelmann, W. H. (1980). An
analysis of the mechanical disadvantage of myocardial infarction in the canine left ventricle. Circ. Res.
47:728-741.
Brutsaert, D. L., and Sys, S. U. (1989). Relaxation and diastole of the heart. Physiol. Rev. 69(4): 1228-315.
Chadwick, R. S. (1982). Mechanics of the left ventricle. Biophys. J. 39(3):279-288.
Costa, K. D., Hunter, P. J., Rogers, J. M., Guccione, J. M., Waldman, L. K., and McCulloch, A. D. (1996a).
A three-dimensional finite element method for large elastic deformations of ventricular myocardium:
!-Cylindrical and spherical polar coordinates. J. Biomech. Engr. 118(4):452-463.
Costa, K. D., Hunter, P. J., Wayne, J. S., Waldman, L. K., Guccione, J. M., and McCulloch, A. D. (1996b).
A three-dimensional finite element method for large elastic deformations of ventricular myocardium:
II-Prolate spheroidal coordinates. J. Biomech. Engr. 118(4):464-472.
de Tombe, P. P., and ter Keurs, H. E. (1992). An internal viscous element limits unloaded velocity of
sarcomere shortening in rat myocardium. J. Physiol. 454:619-42.
Derner, L. L., and Yin, F. C. ( 1983). Passive biaxial mechanical properties of isolated canine myocardium.
J. Physiol. 339:615-630.
Demiray, H. (1976). Large deformation analysis of some basic problems in biophysics. Bull. Math. Biol.
38(6):701-712.
Dokos, S., LeGrice, I. J., Smaill, B. H., Kar, J., and Young, A. A. (2000). A triaxial-measurement shear-test
device for soft biological tissues. J. Biomech. Engr. 122(5):471-8.
Fung, Y. C. (1981). Biomechanics: Mechanical Properties of Living Tissues. New York: Springer-Verlag.
Gaasch, W. H., and LeWinter, M. M. (1994). Left Ventricular Diastolic Dysfunction and Heart Failure.
Philadelphia: Lea & Febiger.
Gilbert, J. C., and Glantz, S. A. (1989). Determinants of left ventricular filling and of the diastolic pressure-
volume relation. Circ. Res. 64(5):827-52.
Glantz, S. A., and Parmley, W. W. (1978). Factors which affect the diastolic pressure-volume curve. Circ.
Res. 42(2):171-80.
Glantz, S. A., Misbach, G. A., Moores, W. Y., Mathey, D. G., Lekven, J., Stowe, D. F., Parmley, W. W.,
and Tyberg, J. V. (1978). The pericardium substantially affects the left ventricular diastolic pressure-
volume relationship in the dog. Circ. Res. 42(3):433-41.
Glass, L., Hunter, P. J., and McCulloch, A. D. (1991). Theory of Heart: Biomechanics, Biophysics and
Nonlinear Dynamics of Cardiac Function. New York: Springer-Verlag.
Guccione, J. M., and McCulloch, A. D. (1993). Mechanics of active contraction in cardiac muscle: Part
!-Constitutive relations for fiber stress that describe deactivation. J. Biomech. Engr. 115(1):72-81.
Guccione, J. M., McCulloch, A. D., and Waldman, L. K. (1991). Passive material properties of intact
ventricular myocardium determined from a cylindrical model. J. Biomech. Engr. 113(1):42-55.
Gupta, K. B., Ratcliffe, M. B., Fallert, M.A., Edmunds Jr., L. H., and Bogen, D. K. (1994). Changes in
passive mechanical stiffness of myocardial tissue with aneurysm formation. Circ. 89(5):2315-26.
Hill, A. V. (1938). Time heart of shortenning and the dynamic constants of muscle. Proc. Roy. Soc.
126:136-195.
Hill, A. V. (1970). First and LastExperiments in Muscle Mechanics. Cambridge: University Press.
Hill, J. M. ( 1973). Partial solutions of finite elasticity - three dimensional deformations. J. Appl. Math.
Phys. 24:609-618.
Holmes, J. W., Yamashita, H., Waldman, L. K., and Covell, J. W. (1994). Scar remodeling and transmural
deformation after infarction in the pig. Circ. 90(1):411-420.
Humphrey, J.D., and Yin, F. C. (1987). A new constitutive formulation for characterizing the mechanical
behavior of soft tissues. Biophys. J. 52( 4 ):563-70.
Humphrey, J. D., Strumpf, R. K., and Yin, F. C. (1990a). Determination of a constitutive relation for
passive myocardium: II. Parameter estimation. J. Biomech. Engr. 112(3):340-6.
Humphrey, J. D., Strumpf, R. K., and Yin, F. C. (1990b). Determination of a constitutive relation for
passive myocardium: I. A new functional form. J. Biomech. Engr. 112(3):333-9.
Hunter, P. J., McCulloch, A. D., Nielsen, P. M. F., and Smail!, B. H. ( 1988). A finite element model of
passive ventricular mechanics. In Spilker, R. L., and Simon, B. R., eds., Computational Methods in
Bioengineering, volume 9. Chicago: ASME. 387-397.
Hunter, P. J., McCulloch, A. D., and ter Keurs, H. E. (1998). Modelling the mechanical properties of
cardiac muscle. Prog. Biophys. Mol. Bioi. 69(2-3):289-331.
Huxley, A. F. (1957). Muscle structure and theories of contraction. Prog. Biophys. Chern. 7:255-318.
Ingels Jr., N. B., Daughters 2nd, G. T., Stinson, E. B., and Alderman, E. L. (1975). Measurement of
midwall myocardial dynamics in intact man by radiography of surgically implanted markers. Circ.
52(5):859-67.
Janicki, J. S., and Weber, K. T. (1980). The pericardium and ventricular interaction, distensibility, and
function. Am. J. Physiol. 238(4):H494-503.
Karlon, W. J., Covell, J. W., McCulloch, A. D., Hunter, P. J., and Omens, J. H. (1998). Automated measure-
ment of myofiber disarray in transgenic mice with ventricular expression of ras. Anat. Rec. 252(4):612-
25.
Karlon, W. J., McCulloch, A. D., Covell, J. W., Hunter, P. J., and Omens, J. H. (2000). Regional dysfunction
correlates with myofiber disarray in transgenic mice with ventricular expression of ras. Am. J. Physiol.
Heart Circ. Physiol. 278(3):H898-906.
Karlon, W. J. (1998). Influence of Myocardial Fiber Organization on Ventricular Function. Ph.d., Univer-
sity of California San Diego.
Kentish, J. C., ter Keurs, H. E., Ricciardi, L., Bucx, J. J., and Noble, M. I. (1986). Comparison between the
sarcomere length-force relations of intact and skinned trabeculae from rat right ventricle. Influence of
calcium concentrations on these relations. Circ. Res. 58(6):755-68.
Kruger, G. W., and Pollack, J. H. (1975). Myocardial sarcomere dynamics during isometric contraction. J.
Physiol. 51:627-643.
Landesberg, A., and Sideman, S. (1994). Coupling calcium binding to troponin C and cross-bridge cycling
in skinned cardiac cells. Am. J. Physiol. 266(3 Pt 2):H1260-71.
Landesberg, A., Markhasin, V. S., Beyar, R., and Sideman, S. (1996). Effect of cellular inhomogene-
ity on cardiac tissue mechanics based on intracellular control mechanisms. Am. J. Physiol. 270(3 Pt
2):Hll01-14.
Lee, M. C., Fung, Y. C., Shabetai, R., and LeWinter, M. M. (1987). Biaxial mechanical properties of human
pericardium and canine comparisons. Am. J. Physiol. 253(1 Pt 2):H75-82.
LeGrice, I. J., Smaill, B. H., Chai, L. Z., Edgar, S. G., Gavin, J. B., and Hunter, P. J. (1995a). Laminar
structure of the heart: Ventricular myocyte arrangement and connective tissue architecture in the dog.
Am. J. Physiol. 269(2 Pt 2):H571-82.
LeGrice, I. J., Takayama, Y., and Covell, J. W. ( 1995b ). Transverse shear along myocardial cleavage planes
provides a mechanism for normal systolic wall thickening. Circ. Res. 77:182-193.
Legrice, I. J., Hunter, P. J., and Smaill, B. H. ( 1997). Laminar structure of the heart: A mathematical model.
Am. J. Physiol. 272(5 Pt 2):H2466-76.
Lin, I. E., and Taber, L. A. ( 1994). Mechanical effects oflooping in the embryonic chick heart. J. Biomech.
27(3):311-21.
Lin, D. H., and Yin, F. C. (1998). A multiaxial constitutive law for mammalian left ventricular myocardium
in steady-state barium contracture or tetanus. J. Biomech. Engr. 120(4):504-17.
Maron, B. J., Bonow, R. 0., Cannon 3rd, R. 0., Leon, M. B., and Epstein, S. E. (1987). Hypertrophic
cardiomyopathy. Interrelations of clinical manifestations, pathophysiology, and therapy (1). New Engl.
J. Med. 316(13):780-9.
Maruyama, Y., Ashikawa, K., lsoyama, S., Kanatsuka, H., Ino-Oka, E., and Takishima, T. (1982). Mechan-
ical interactions between four heart chambers with and without the pericardium in canine hearts. Circ.
Res. 50(1):86-100.
May-Newman, K., Omens, J. H., Pavelec, R. S., and McCulloch, A. D. (1994). Three-dimensional trans-
mural mechanical interaction between the coronary vasculature and passive myocardium in the dog.
Circ. Res. 74( 6): 1166-78.
Mazhari, R., Omens, J. H., Covell, J. W., and McCulloch, A. D. (2000). Structural basis of regional
dysfunction in acutely ischemic myocardium. Cardiovasc. Res. 47(2):284-93.
ter Keurs, H. E., Rijnsburger, W. H., van Heuningen, R., and Nagelsmit, M. J. (1980). Tension development
and sarcomere length in rat cardiac trabeculae. Evidence of length-dependent activation. Circ. Res.
46(5):703-14.
McCulloch, A. D., and Omens, J. H. (1991). Non-homogeneous analysis of three-dimensional transmural
finite deformation in canine ventricular myocardium. J. Biomech. 24(7):539-48.
McCulloch, A. D., Smaill, B. H., and Hunter, P. J. (1989). Regional left ventricular epicardial deformation
in the passive dog heart. Circ. Res. 64:721-733.
McLean, M., Ross, M. A., and Prothero, J. (1989). Three-dimensional reconstruction of the myofiber
pattern in the fetal and neonatal mouse heart. Anat. Rec. 224(3):392-406.
Meier, G. D., Bove, A. A., Santamore, W. P., and Lynch, P. R. ( 1980a). Contractile function in canine right
ventricle. Am. J. Physiol. 239(6):H794-804.
Meier, G. D., Ziskin, M. C., Santamore, W. P., and Bove, A. A. (1980b). Kinematics of the beating heart.
IEEE Trans. Biomed. Engr. 27(6):319-29.
Mirsky, I., and Rankin, J. S. (1979). The effects of geometry, elasticity, and external pressures on the
diastolic pressure-volume and stiffness-stress relations. How important is the pericardium? Circ. Res.
44(5):601-11.
Mirsky, I. (1976). Assessment of passive elastic stiffness of cardiac muscle: Mathematical concepts, physi-
ologic and clinical considerations, directions of future research. Prog. Cardiovasc. Dis. 18(4):277-308.
Nevo, E., and Lanir, Y. (1989). Structural finite deformation model of the left ventricle during diastole and
systole. J. Biomech. Engr. 111(4):342-9.
Nielsen, P.M., Grice, I. J. L., Smail!, B. H., and Hunter, P. J. (1991). Mathematical model of geometry and
fibrous structure of the heart. Am. J. Physiol. 260( 4 Pt 2):H 1365-78.
Novak, V. P., Yin, F. C., and Humphrey, J. D. (1994). Regional mechanical properties of passive my-
ocardium. J. Biomech. 27(4):403-12.
O'Dell, W. G., and McCulloch, A. D. (2000). Imaging three-dimensional cardiac function. Ann. Rev.
Biomed. Eng. 2:431-56.
Oden, J. T. (1972). Finite Elements of Nonlinear Continua. New York: McGraw-Hill Book Co.
Ogden, R. W. (1984). Non-Linear Elastic Deformations. New York: Ellis Horwood.
Omens, J. H., May, K. D., and McCulloch, A. D. (1991). Transmural distribution of three-dimensional
strain in the isolated arrested canine left ventricle. Am. J. Physiol. 261(3 Pt 2):H918-28.
Omens, J. H., MacKenna, D. A., and McCulloch, A. D. (1993). Measurement of strain and analysis of
stress in resting rat left ventricular myocardium. J. Biomech. 26(6):665-76.
Panerai, R. B. (1980). A model of cardiac muscle mechanics and energetics. J. Biomech. 13(11):929---40.
Patterson, S. W., and Starling, E. H. (1914). On the mechanical factors which determine the output of the
ventricles. J. Physiol. 48:357.
Peng, S. H., and Chang, W. V. (1997). A compressible approach in finite element analysis of rubber-elastic
materials. Comput. & Structures 62:573-593.
Ruegg, J. C. (1988). Calcium in Muscle Activation: A Comparative Approach. Berlin: Springer-Verlag.
Saffitz, J. E., Kanter, H. L., Green, K. G., Tolley, T. K., and Beyer, E. C. (1994). Tissue-specific deter-
minants of anisotropic conduction velocity in canine atrial and ventricular myocardium. Circ. Res.
74(6):1065-70.
Sagawa, K. ( 1988). Cardiac Contraction and the Pressure- Volume Relationship. New York: Oxford
University Press.
Salisbury, P. F., Cross, C. E., and Rieben, P. A. (1960). Influence of coronary artery pressure upon myocar-
dial elasticity. Circ. Res. 8:794.
Schmid, P., Stuber, M., Boesiger, P., Hess, 0. M., and Niederer, P. (1995). Determination of displacement,
stress- and strain-distribution in the human heart: A FE-model on the basis of MR imaging. Techno[.
Health Care 3(3):209-14.
Schoenberg, M. (1980). Geometrical factors influencing muscle force development. II. Radial forces.
Biophys. J. 30(1):69-77.
Simpson, G., Fisher, C., and Wright, D. K. (2001). Modeling the interactions between a prosthetic socket,
polyurethane liners and the residual limb in transtibial amputees using non-linear finite element analy-
sis. Biomed. Sci. Instrum. 37:343-7.
Slinker, B. K., and Glantz, S. A. (1986). End-systolic and end-diastolic ventricular interaction. Am. J.
Physiol. 251(5 Pt 2):Hl062-75.
Sonnenblick, E. H., Napolitano, L. M., Daggett, W. M., and Cooper, T. (1967). An intrinsic neuromuscular
basis for mitral valve motion in the dog. Circ. Res. 21(1):9-15.
Spencer, A. J. M. (1980). Continuum Mechanics. London: Lonman Press.
Spotnitz, H. M., Spotnitz, W. D., Cottrell, T. S., Spiro, D., and Sonnenblick, E. H. (1974). Cellular basis
for volume related wall thickness changes in the rat left ventricle. J. Mol. Cell Cardiol. 6(4):317-31.
Streeter Jr., D. D., and Hanna, W. T. (1973). Engineering mechanics for successive states in canine left
ventricular myocardium. I. Cavity and wall geometry. Circ. Res. 33(6):639-55.
Streeter Jr., D. D., Spotnitz, H. M., Patel, D.P., Ross Jr., J., and Sonnenblick, E. H. ( 1969). Fiber orientation
in the canine left ventricle during diastole and systole. Circ. Res. 24(3):339---47.
Streeter Jr., D. D. (1979). Gross morphology and fiber geometry of the heart. In Bethesda, M.D., ed.,
Handbook of Physiology. American Physiological Society. 61.
Taber, L.A. (1991). On a nonlinear theory for muscle shells: Part II-Application to the beating left ventri-
cle. J. Biomech. Engr. 113(1):63-71.
Torrent-Guasp, F. (1973). The Cardiac Muscle. Madrid: Juan March Foundation.
Tozeren, A. ( 1985). Continuum rheology of muscle contraction and its application to cardiac contractility.
Biophys. J. 47(3):303-9.
Usyk, T. P., Mazhari, R., and McCulloch, A. D. (2000). Effect of laminar orthotropic myofiber architecture
on regional stress and strain in the canine left ventricle. J. Elasticity 61(1/3):143-164.
Usyk, T. P., Omens, J. H., and McCulloch, A. D. (2001 ). Regional septal dysfunction in a three-dimensional
computational model of focal myofiber disarray. Am. J. Physiol. Heart Circ. Physiol. 281 (2):H506--14.
Vetter, F., and McCulloch, A. D. (1998). Three-dimensional analysis of regional cardiac function: A model
of the rabbit ventricular anatomy. Pro g. Biophys. Mol. Biol. 69:157.
Vetter, F. J., and McCulloch, A. D. (2000). Three-dimensional stress and strain in passive rabbit left
ventricle: A model study. Ann. Biomed. Engr. 28(7):781-92.
Waldman, L. K., Fung, Y. C., and Covell, J. W. (1985). Transmural myocardial deformation in the canine
left ventricle. Normal in vivo three-dimensional finite strains. Circ. Res. 57(1 ): 152-63.
Weis, S. M., Omens, J. H., and McCulloch, A. D. (2001). Ventricular tissue adaptation associated with
collagen deficiency in the osteogenesis imperfecta murine. J. Biomech. Engr. in press.
Wong, A. Y. (1971). Mechanics of cardiac muscle, based on Huxley's model: Mathematical stimulation of
isometric contraction. J. Biomech. 4(6):529-40.
Wong, A. Y. (1972). Mechanics of cardiac muscle, based on Huxley's model: Simulation of active state
and force-velocity relation. J. Biomech. 5(1):107-17.
Yellin, E. L., Hori, M., and Yoran, C. (1986). Left ventricular relaxation in the filling and nonfilling intact
canine heart. Am. J. Physiol. 250:620.
Yin, F. C., Chan, C. C., and Judd, R. M. ( 1996). Compressibility of perfused passive myocardium. Am. J.
Physiol. 271(5 Pt 2):Hl864-70.
Young, A. A., and Axel, L. ( 1992). Three-dimensional motion and deformation of the heart
wall: Estimation with spatial modulation of magnetization - a model-based approach. Radiology
185(1):241-7.

Biomechanics of Vasc Soft Tissue

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Biomechanics of Vasc Soft Tissue

Caricato da

Copyright:

Formati disponibili

CISM COURSES AND LECTURES

The Secretary General

Former Secretary General

The series presents lecture notes, monographs, edited works and

COURSES AND LECTURES- No. 441

BIOMECHANICS OF SOFT TISSUE

t Springer-Verlag Wien GmbH

This work is subject to copyright.

ISBN 978-3-211-00455-5 ISBN 978-3-7091-2736-0 (eBook)

In order to make this volume available as economically and as

The Cardiovascular System -Anatomy, Physiology and Cell Biology

Mechanical Properties of Soft Tissues and Arterial Walls

Nonlinear Elasticity, Anisotropy, Material Stability and Residual Stresses in

Intracranial Saccular Aneurysms

Remodeling of Arteries in Response to Changes in their Mechanical Environ-

Computational Methods for Soft Tissue Biomechanics

J.D. Humphreyl and A.D. McCulloch 2

1.1 Basic Anatomy

1.2 Histology and Cell Biology

1.3 Pressure-Volume Behavior

P(t) = E(t)[v(t) - V], (1)

PVA = E+Il. (3)

2.1 The Vascular Tree

2.3 Physiology and Cell Biology

3 Some Common Diseases

Graduate School of Engineering Science, Osaka University

Abstract. Mechanical properties of biological tissues are fundamental and prerequisite

1 Mechanical Properties of Biological Soft Tissues

1.1 Test Methods

Nuchal Sole Intestinal

0 0.25 0.5 0 0.04 0.08 0.12 0 0.25 0.5

1.2 Basic Properties Unique to Soft Tissues

Figure 3. A model of biological soft tissue.

Rabbit ventral skin

Extension ratio )..x,).. y

Bovine articular cartilage

25~----~----~1------~----~1 ------~----~1 ------~--~

5. Strain Rate Insensitivity

Canine lung parenchyma

CT = Load/unload time (sec)

1.1 1.2 1.3 1.4 1.5 1.6 1.7

1.3 Mathematical Description of Mechanical Behavior

In a simple tensile test, the only non-zero component of stress, denoted by S, is

Human dura mater and I

0 0.5 1.0 1.5 2.0

is given by Fung (1973) in the form

+a3Ei2 + a3E~ 1 + 2a4EnE22 + 11E~ 1 + 12E~2 + r4Ei 1 E22 + r5EnE~2 ), (9)

Different blood vessels

0 0.05 0.10 0.15 0.20 0.25 0.30

where A 1 , A 2 , and X are functions of the strain components.

Maxwell model Voigt model Kelvin model

F,u F,u F,u

Figure 11. Models for viscoelastic materials.

Thus, the Maxwell model is formulated as

and the Kelvin model is described by

Frog sartorius muscle

1.4 Contraction of Muscle

2 Mechanical Properties of Arterial Walls

TA = Thoracic aorta, AA = Abdominal aorta, CA = Carotid artery, FA = Femoral artery, AL =

I Location A TA AA I CIA I FA CCA PA

R = Radius, Ep = Pressure strain elastic modulus, A = Ascending aorta, TA

2.1 Arterial Structure and Composition

2.2 Pressure-Diameter Relation

Experimental system for arterial pressure-diameter relations

'-------~·--· L_ ~c_-- ·1~~~~ ~L_~~ _j_--~-- _]___