281

Finite element analysis in spine research

M J Fagan1*, S Julian1 and A M Mohsen2
1Medical Engineering, Department of Engineering, University of Hull, Hull, UK
2Department of Orthopaedics and Traumatology, Hull Royal In rmary, Hull, UK

Abstract: Finite element analysis is a widely accepted tool used in many industries and research
activities. It allows new designs to be thoroughly tested before a prototype is even manufactured,
components and systems which cannot readily be experimented upon to be examined, and diagnostic
investigations to be undertaken.
Finite element models are already making an important contribution to our understanding of the
spine and its components. Models are being used to reveal the biomechanical function of the spine
and its behaviour when healthy, diseased or damaged. They are also providing support in the design
and application of spinal instrumentation.
The spine is a very complex structure, and many of the models are simpli ed and idealized because
of the complexity and uncertainty in the geometry, material properties and boundary conditions of
these problems. This type of modelling simpli cation is not peculiar to spinal modelling problems.
Indeed, the idealization is often a strength when there is such uncertainty and variation between one
individual and another, allowing causeeVect relationships to be isolated and fully explored, and the
inherent variability of experimental tests to be eliminated.
This paper reviews the development of nite element analysis in spinal modelling. It shows how
modelling provides a wealth of information on our physiological performance, reduces our dependence on animal and cadaveric experiments and is an invaluable complement to clinical studies. It
also leads to the conclusion that, as computing power and software capabilities increase, it is quite
conceivable that in the future it will be possible to generate patient-speci c models that could be used
for patient assessment and even pre- and inter-operative planning.
Keywords: nite element analysis, spine, modelling

INTRODUCTION

Finite element analysis is an essential part of many of

todays engineering activities. It is a very well-tried and
tested technique used in the design of numerous products
from electronic systems and domestic appliances through
to cars, aircraft and large civil engineering constructions.
The method was rst developed in the 1950s in the aircraft industry and continues to be an indispensable tool
in the design of most of the critical components in
todays aircraft. The rst application of nite element
analysis in biomechanics was probably reported in 1972
by Brekelmans et al. [1]. Since then the number of applications has grown enormously with the promise of a
detailed insight into the behaviour of biological and
composite biologicalbiomechanical systems.
The MS was received on 23 April 2002 and was accepted after revision
for publication on 1 May 2002.
* Corresponding author: Department of Engineering, University of Hull,
Cottingham Road, Hull HU6 7RX, UK.
H03402 IMechE 2002

As available computing power has increased over the

years, so more complex problems have been analysed
and investigated, including the spine and its components
There are four ways in which nite element modelling
can be used in spine research, namely
(a) to provide an assessment of the spine in health,
(b) to provide an assessment of the spine as altered by
disease, degeneration, ageing, trauma or surgery,
(c) to provide an assessment of the spine with spinal
instrumentation and
(d ) to assist in the design and development of that spinal
instrumentation.
Experimental analysis and clinical studies can also
provide this type of information, but modelling provides
us with information that laboratory experiments cannot.
For example, it can predict the value of stresses in the
discs and vertebrae and very detailed motion data. It
also allows exploration of an essentially limitless range
of physiological conditions and combinations of conditions which helps the biomechanics of the system to
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

282

M J FAGAN, S JULIAN AND A M MOHSEN

be understood. Similarly, nite element analysis permits

exploration of a limitless range of diVerent treatment
options and surgical interventions and allows testing
and optimization of any prostheses before it is even
manufactured.
The spine is a very complicated part of our skeleton;
it consists of a number of complex components which
have elaborate structures and non-linear material
properties. However, this paper does not discuss or reference the physiology of the spine, which is very well
described elsewhere in the literature.
The following review of nite element modelling of
the spine attempts to collect similar models together into
a number of sections, namely
(a)
(b)
(c )
(d )
(e)
(f )

whole spine (simple) models,

vertebral body models,
intervertebral disc and motion segment models,
lumbar spine,
cervical spine models and
models representing spinal mobilitycondition and
instrumentation.

However, there is invariably an overlap in these sections and the design and application of the models. In
essence, nite element spine models are grouped into
those which deal with non-pathological regional anatomy ( lumbar, thoracic, cervical and whole spine), nonpathological organ anatomy (vertebral body, disc and
motion segment), pathological anatomy conditions and
treatment of spinal pathology. Also, the various research
groups have often started with relatively simple models
of individual components of the spine and extended and
developed the models to the point where larger sections
of the spine are represented. Hence their research eVorts
will frequently be divided between a number of the
sections described below.
2

nite element model of the spine to evaluate the mechanical response of a pilot being ejected from a planes cockpit ( Fig. 1). The model consisted of the vertebrae, pelvis,
head, ribs from T1 to T10, upper and lower viscera,
ligaments, muscles and connective tissue. It was assumed
that all the skeletal components were rigid bodies, with
each vertebra being modelled by only one rigid element.
Spring elements were used for the ligaments and harness
restraints, and beam elements for the intervertebral discs.
The model was subjected to a non-linear force applied
to the base of the model (the seat) to simulate ejection.
Their model was later improved to represent more accurately the articular facets and the muscle groups in the
neck [7] and used to model the cervical spine under
lateral and frontal impact accelerations.
A similar model was developed by Roberts and Chen
[8 ] who used their model to examine the eVects of a car
crash on the sternum. The same model data were used
later by Sundaram and Feng [9] to develop a model with
simple representations of some internal organs and again
under a number of front impact conditions.
A simpli ed three-dimensional nite element model
was developed by Dietrich et al. [10] to look at the
behaviour of the complete spine and is shown in Fig. 2.
It included intervertebral discs that were modelled with

WHOLE SPINE (SIMPLE ) MODELS

These so-called simple models are those which use just

a few (usually one or three) elements to represent each
motion segment, but simulate a large section of the spine.
The rst analytical model of the spine was probably published as early as 1957 for pilot ejection studies and used
a simple springmass con guration [2]. Increasingly
complex springmass models were then developed,
including dampers to represent the intervertebral discs
[3]. It was not until 1970 that a model with more realistic
disc properties was proposed by Orne and Liu [4]. Their
model included discs with shear and bending resistance
that used a three-parameter viscoelastic forcede ection
relationship which varied with vertebral position. The
model was subsequently extended by Prasad and King
[5] to include articular facet interactions by the use of
spring elements.
Belytschko et al. [6 ] developed a three-dimensional
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

Fig. 1

Lateral view of a simpli ed model to evaluate the

mechanical response of a pilot being ejected from an
aircraft. (From Belytschko et al. [6 ])
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

Fig. 3

Fig. 2

A three-dimensional model of the spine and ribcage

that included a simple disc model and some muscle
and ligament detail. (From Dietrich et al. [10])

anisotropic elastic properties for the annulus and an

incompressible space for the nucleus pulposus. All bone
components were again assumed to be rigid, but both
ligament and muscle eVects were included. The model
was used to estimate the forces and stresses in the diVerent elements for diVerent positions of the spine [11].
Another model of a spine, ribcage and pelvis has been
used to predict the vertebral displacements due to manipulative techniques used during assessment and treatment
of spinal disorders [12] (Fig. 3). In this model the intervertebral discs were modelled by beam elements, and it
was assumed that the patient was relaxed and hence the
muscle forces were negligible.
A few more of these simple spine models which
have been developed to simulate conditions such as
scoliosis are dealt with in the last section describing the
diVerent applications of nite element analysis in
instrumentation.
3

VERTEBRAL BODY MODELS

The models in this and subsequent sections use many

nite elements to model each component and hence provide much more information on the performance of each
part, including the internal stress and strains, and result
in a more complex behaviour. For example, Hakim and
H03402 IMechE 2002

283

Model of a spine, ribcage and pelvis used to predict

responses to lumbar manipulative forces. This model
used short beam elements to represent the discs. (From
Lee et al. [12])

King [13, 14] developed a three-dimensional nite

element model of an isolated vertebra with anatomical
details of the bone obtained through sectioning and
direct measurement. Even so the number of elements
representing the vertebra was less than 150, as shown in
Fig. 4. Some comparison was made with experimental
data and it was observed that signi cant discrepancies
existed because of the very complex nature of the problem and uncertainties in the geometry and material
properties of the bone. Hakim and King were probably
the rst to consider the facet joints in their model, but
the disc was represented by simple linear axial elements.
The model was extended later to a full motion segment
by including a disc with linear homogeneous isotropic
properties [15].
Component geometries may be obtained either
through the use of coordinate-measuring machines or
three-dimensional digitizers, or directly from computer
tomography (CT ) scan images. For example, Bozic et al.
[16 ] developed a solid three-dimensional nite element
model of a cervical vertebra from a CT image to investigate burst fracture mechanisms. Each voxel of the scan
data was converted directly to a cube-shaped nite
element of size 1.25 mm, as illustrated in Fig. 5. The
advantage of this approach is that the apparent bone
density of each voxel can be obtained from the CT data,
and Youngs modulus and strength of each element then
varied accordingly (usually by empirical relationships of
the type derived by Carter and Hayes [17]). The 8590
elements that were used to model the vertebra were supported by springs to represent the adjacent discs. A much
more geometrically accurate model of the rst cervical
vertebra (atlas) is shown in Fig. 6. This was produced
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

284

M J FAGAN, S JULIAN AND A M MOHSEN

Fig. 4

One of the rst three-dimensional models of a human vertebra, produced by Hakim and King with
anatomical details obtained by direct measurement. Only half the body was represented because of
planar symmetry, but the model did include the posterior elements. (From Hakim and King [13])

by a three-dimensional digitizer and was used to

investigate fracture mechanisms of the bone [18].
The strength of an isolated L1 vertebral body was also
examined by Whyne et al. [19]. They used a much simpli ed representation of the geometry and material properties (shown in Fig. 7) to examine the contribution of
the pedicles and posterior arch to the vertebral body
strength. Mizrahi et al. [20] considered the eVects of
osteoporosis on an idealized L3 vertebral body by
decreasing Youngs modulus of the bone and examining
the changes in stress when a compressive load was
applied to the model. The material properties varied
through their model as a function of (QCT ) density data.
Their analyses showed how central end-plate and transverse fractures might initiate from a simple uniform
compression of the end plate.
The load-bearing roles of the shell and core of a vertebral body have been examined by a number of
researchers. Silva et al. [21] examined the eVects of shell
thickness and material property values of the cortical
and trabecular bone. They concluded that the shell
accounted for only approximately 10 per cent of the
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

vertebral strength, with the trabecular core being the key

structural component of vertebral bodies. A non-linear
microstructural foam-based model of trabecular bone
was used by Overaker et al. [22] in their examination of
vertebral body mechanics. They concluded that the ratio
of the load carried by the shell to that carried by the
trabecular core ranged from 38 to 83 per cent depending
on age (i.e. porosity of the core) and curvature of the
lateral wall of the vertebra. Other models have been
developed to examine very particular aspects of the vertebral body. For example, Inoue et al. [23] devised a
model with accurately discretized posterior elements so
that the isthmus stresses of the lower lumbar neural arch
could be examined in detail.
Bone adaptive remodelling of vertebra has been
undertaken by Goel et al. [24] to demonstrate how the
concave pro le of a vertebral body can develop from an
initial cylindrical shape. The model included the spinal
ligaments and assumed all the material properties to be
isotropic throughout. The vertebrae adapted by minimization of the strain energy density through movement
of the external nodes and alteration of the modulus of
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

Fig. 7

Fig. 5

Finite element model of a C4 vertebra obtained directly

from CT data, where each voxel was converted directly
into a nite element. (From Bozic et al. [16 ])

Fig. 6

Finite element model of the rst cervical vertebra

(atlas) developed by direct measurement of a cadaveric
specimen and used to examine fracture mechanisms.
(From Teo et al. [18])

the internal elements, so that its shape and apparent

density (re ected by Youngs modulus) were consistent
with experimental observations.
The most accurate way of specifying the spine models
geometry is to input the data directly from CT images
[25]. There are now software packages [26 ] which can
convert the data from CT and magnetic resonance imaging (MRI ) scans into a format that can be interpreted
by engineering computer aided draughting software,
which can then be input into commercial nite element
software. Clearly, the dimensions of much of the verH03402 IMechE 2002

285

Simpli ed model of an isolated L1 vertebral body used

to examine the eVects of the pedicles and posterior arch
on vertebral strength. (From Whyne et al. [19])

tebrae are not critical to most modelling applications,

provided that the primary features reliably re ect their
function and properties. Although the secondary features must still be similar if the attachment and action
of the muscles and ligaments are to be reasonable. For
example, Fig. 8 shows a geometrically accurate model of
a vertebra produced by Fagan et al. [27] generated from
CT data, while Fig. 9 shows a simpli ed version of the
same data included in a model of two-motion segments.
Stepney et al. [28] developed a parametric solid model
of the geometry where the basic geometric information
was taken from a single spine and each vertebra was
de ned by 60 parameters, while each disc required just
six. However, while they generated the basic model, they
do not appear to have used it to undertake a nite
element study with it. A similar approach was used by
Lavaste et al. [29] and Robin et al. [30] a few years

Fig. 8

A geometrically accurate model of a vertebra generated

directly from CT data [27]
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

286

M J FAGAN, S JULIAN AND A M MOHSEN

Fig. 9

A simpli ed nite element model of two motion

segments developed with the same CT data used in the
vertebra model in Fig. 8 [27]

earlier who used just six parameters to de ne their nite

element models of each vertebra.
4

INTERVERTEBRAL DISC AND MOTION

SEGMENT MODELS

Belytschko et al. [31] were the rst to present details of

a nite element analysis of an intervertebral disc and
adjacent vertebrae in 1974. They modelled the problem
assuming axial symmetry with linear orthotropic material properties for the disc, as shown in Fig. 10. The same
axisymmetric model was subsequently extended by
assuming that the annulus had non-linear orthotropic
properties, the actual values of which were derived by
comparison with experimental measurements [32].
Lin et al. [33] developed a three-dimensional model
of the disc but assumed linear orthotropic properties of
the disc, which again were optimized to ensure that the
analytical results matched with experimental data.
In contrast, Spiker [34] reported a parametric study
of a much simpli ed axisymmetric model of a disc
(Fig. 11) with linear isotropic properties used for the
annulus. They considered a range of values for the geometric and material properties to determine which had
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

the most signi cant eVect on the discs response when

under compression. The representation of the annulus
was developed further and modelled as a layered brereinforced composite [35]. The disc was then analysed
under a variety of asymmetric loading conditions
including torsion, shear and bending.
One of the most signi cant contributions to the modelling of intervertebral discs was made by Shirazi-Adl et al.
[36 ] in 1984. They developed a three-dimensional nonlinear nite element model of the L2L3 disc body unit.
The model consisted of cortical bone, cancellous bone,
bony end plate and the intervertebral disc and is illustrated in Fig. 12. The discs annulus was represented as
a composite of collagenous bres embedded in a matrix
of ground substance, while the nucleus was modelled as
an incompressible inviscid uid. The axial elements used
to model the bres were arranged in a criss-cross pattern
around the annulus and de ned with non-linear properties to re ect the softening of the bres at higher strains.
Because the disc experiences large displacements and
strains under normal compressive loading, the analysis
required both a non-linear material and a non-linear
geometry solution. This basic model was developed
further to include contact at the facet joints and analysis
of the motion segment under pure sagittal plane
moments [37], and later under combinations of more
complex loading conditions [38]. Shirazi-Adl also examined the eVects of facet geometry in detail [39] and
undertook a more detailed examination of the constitutive models of the annulus [40], the role of the nucleus
pulposus [41] and the eVects of the bones properties
[42]. Finally, the same model was used by Kasra et al.
[43] to predict free and forced-vibration responses of the
motion segment. Some time later Agroubi and ShiraziAdl [44] modi ed the model again to include poroelastic
properties for the disc.
Rao and Dumas [45] used a similarly constructed
model to that of Shirazi-Adl but assumed a simpli ed,
circular and symmetric geometry ( Fig. 13) and varied
the properties of the constituent materials to examine
their eVects on the discs mechanical behaviour.
The rst poroelastic nite element model of an intervertebral disc was probably reported by Simon et al. [46 ]
in 1985. Poroelasticity is a special case of viscoelasticity
where the uid phase can move with respect to the solid
phase. As a result, when a static load is applied, the uid
is forced out of the disc so that the solid phase supports
the load, with removal of the load then allowing the disc
to regain its shape. In contrast, with a transient shortterm load the disc behaves in a nearly incompressible
manner. The model was extended by Laible et al. [47]
to include the eVects of swelling caused by osmotic pressure such that load bearing by the uid phase may occur
even at equilibrium, with a corresponding reduction in
solid phase support and stresses. Some measurements of
the internal pressures in discs subjected to diVerent loading conditions have been undertaken and shown to give
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

Fig. 10

The rst model of an intervertebral disc and adjacent vertebrae developed by Belytschko et al. in
1974. The model is axisymmetric with a horizontal plane of symmetry. (From Belytschko et al. [31])

good agreement with poroelastic nite element models

[48, 49]. It is expected that these poroelastic models
will give further insight into abnormal disc function,
degeneration and failure.
The time-dependent response of a lumbar intervertebral disc was considered by Wang et al. [50] with the
development of a non-linear viscoelastic motion segment. A useful exploded view of their model is presented
in Fig. 14, revealing the typical construction of a motion
segment model, including vertebral bodies, end plates
and a disc unit. They subsequently used the model to
examine the dynamic response of an L2L3 motion
segment in cyclic axial compressive loading [51] and
to determine the eVects of diVerent rates of sagittal
exion [52].
A number of researchers have speci cally considered
failure of the disc. For example, Ueno and Liu [53] generated a three-dimensional solid nite element model of
a motion segment with geometric non-linearity to study
the eVects of torsion as a cause of failure in the lumbar
intervertebral joint. They found that disc failure was
most likely on the lateral aspect, but that posterior or
H03402 IMechE 2002

287

posterolateral disc rupture was unlikely to be caused by

torsion. The eVects of bending, torsion and diurnal uid
changes in the disc were considered by Lu et al. [54, 55]
who con rmed that disc prolapse was more likely under
combined loading and inclined to occur at the posterior inner annulus at the junction of the disc and end
plate. They also reported that failure was more likely
with a saturated disc in the morning after the spine has
been in a relatively unloaded state through the night.
Interlaminar shear stresses and laminae separation were
examined by Goel et al. [56 ] for L3L4 under simple
axial compression. They concluded that circumferential
tears were most likely to occur in the posterolateral
regions of the disc. Natarajan and Andersson [57] examined the eVect of disc height in relation to cross-sectional
area and showed that, the greater the heightarea ratio,
then the higher is the risk of disc failure.
Other researchers have also presented their work on
modelling the intervertebral disc and lumbar motion segments [58, 59] and used the models to examine the eVects
of the components geometries and material properties.
For example, the eVect of disc height was considered by
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

288

M J FAGAN, S JULIAN AND A M MOHSEN

Fig. 11

A simple axisymmetric model of an intervertebral

disc used to determine equivalent bre composite
properties for the annulus. (From Spiker et al. [35])

Lu et al. [60], who concluded that it had a signi cant

eVect on the deformation and stress levels of the disc
but did not eVect the intradiscal pressure.
Many models of individual motion segments include
ligaments around the motion segment. Cable elements
are usually used, because they, like the ligaments, can
only support tensile loads; see for example Fig. 15 which
is from Lu et al. [54] and shows details of the posterior
ligaments. In particular, the models include the anterior
and posterior longitudinal ligaments, the intertransverse
and interspinous ligaments, the capsular ligament, the
ligamentum avum and supraspinous ligament. Goel
et al. [61] were probably the rst to consider the eVects
of muscle forces. They derived the muscle forces across
the L3L4 segment by a separate optimization technique
(for a person undergoing a speci ed lifting action) and
applied those forces to a ligamentous motion segment
model. The results indicated that the muscles eVected
the translation and rotation of the motion segment as
expected, but speci cally decreased the intradiscal pressure while increasing the load-bearing of the facet joints.
Figure 16 shows the complexity of their model as
reported in Kong et al. [62] and extended to two motion
segments.
5

LUMBAR SPINE MODELS

The ligamentous motion segment model of Shirazi-Adl

et al. [37] was used as the basic component in a number
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

of three-dimensional models of the lumbar spine. Some

details of how such models were generated directly from
CT data were reported by Breau et al. [25]. These models
have been used to undertake a very thorough analysis
of the biomechanics of the lumbar spine, its movement
and stability under a variety of loading modalities
[6368].
Lavaste et al. [29] developed a parametrized threedimensional nite element model of the lumbar spine.
They used just six parameters to de ne the geometry of
each vertebra, where the parameters were obtained by
digitizing two X-rays; (anteroposterior and lateral ). The
parameters were width, depth, height and concavity of
the vertebral body, and the total height and depth of the
vertebra, where the remainder of the geometry was simpli ed to regular shapes. The in uence of these geometrical parameters on the behaviour of the lumbar spine was
then subsequently examined [30].
The motion segment model reported by Goel et al.
[61], which included muscle forces, was also developed
into a model of the thoracolumbar spine by Kong et al.
[62], where forces in the muscles of the lumbar region
were considered. They used the model to predict the
behaviour of the spine and load transfer paths through
it during static lifting in the sagittal plane with normal
and dysfunctional muscle activity. They concluded that
muscle dysfunction (i.e. reduced muscle activity) destabilized the spine, reducing the role of the facet joints in
transmitting loads, and shifted loads to the discs and
ligaments.
6

CERVICAL SPINE MODELS

Cervical spine modelling is primarily of interest to clinicians and biomechanical engineers from the point of
view of cervical spinal cord injury, and to date there has
been much less eVort spent on modelling the cervical
spine and its components than the lumbar spine.
The rst three-dimensional model of the cervical spine
(headT1) was probably reported by Kleinberger [69]
in 1993 and is shown in Fig. 17. It was a relatively crude
model in particular with respect to the much simpli ed
geometry. Voo et al. [70] used CT images at 1.0 mm
intervals in order to generate the geometry of their
C4C6 segment of the spine accurately. Linear isotropic
and homogeneous properties were assumed for all the
components including the discs, but parametric studies
of the eVects of the material properties were conducted
later [71, 72]; Fig. 18 shows a view of the model used in
these papers [72].
Maurel et al. [73] presented a simpli ed parametric
nite element model of the lower cervical spine and used
the model to compare the characteristics of 53 lower
cervical vertebrae, whose key geometric features were
obtained by a coordinate-measuring machine. In particular they examined the in uence of the posterior
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

Fig. 12

Fig. 13

289

The three-dimensional non-linear model of the L2L3 disc body unit developed by Shirazi-Adl et al.
in 1983. Because of symmetry, only a quarter of the joint was modelled, with symmetry about the
sagittal plane and mid-horizontal plane. (From Shirazi-Adl et al. [36 ])

A simpli ed circular symmetric model of a disc used

to examine the eVect of the material properties on the
discs behaviour. (From Rao and Dumas [45])

articular facets which are observed to vary signi cantly

between individuals.
An isolated cervical spine facet joint capsule was also
examined in detail by Kumaresan et al. [74]. They
H03402 IMechE 2002

Fig. 14

An exploded view of a typical motion segment model,

showing the vertebrae, end plates and intervertebral
disc. (From Wang et al. [50])
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

290

Fig. 15

Fig. 16

M J FAGAN, S JULIAN AND A M MOHSEN

A full three-dimensional nite element model of the

L2L3 motion segment showing typical ligament
attachments included in many of todays models.
(From Lu et al. [54])

Finite element model of two motion segments from

the lumbar spine which includes both ligaments and
muscles. (From Kong et al. [62])

Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

Fig. 17

Simpli ed nite element model of the cervical spine

developed by Kleinberger to investigate injuries
arising from car crashes. (From Kleinberger [69])

Fig. 18

Geometrically accurate model of lower cervical

spine (C4C6) generated from 1 mm slices. (From
Kumaresan et al. [72])
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

modelled the joint using four diVerent non-linear

approaches. In the rst two, slide-line elements and contact elements were used to represent the synovial uid
between the articulating cartilages. In the hyperelastic
model, incompressible solid elements approximated the
synovial uid while, in the uid model, hydrostatic uid
elements were used. They considered compression,
exion, extension and lateral bending and concluded
that the uid element best represented the actual facet
joint anatomy and behaviour. Goel and Clausen [75]
also studied the load path through the C5C6 cervical
motion segment under various loading modalities.
7

MODELLING OF SPINAL CONDITIONS AND

INSTRUMENTATION

Apart from using the nite element models to broaden

our understanding of the way in which the spine and its
components function, the models can also most usefully
be used to simulate diVerent spinal conditions and their
treatment, and to assist in the design and analysis of new
spinal instrumentation. At the present most of this modelling is not patient speci c, but there is the likelihood
that patient-speci c models will be used to explore the
likely eVects of diVerent treatments and possibly as a
tool to assist in the more fundamental process of
diagnosis of spinal problems [76].
Lotz et al. [77] studied the degeneration of the intervertebral disc with compressive loads by nite element
analysis and experimental measurement with an in vivo
mouse model. They found that the disc degenerated with
increasing loads in a mode suggested by the stress distribution from the nite element model, but some structure
and function was regained in some cases after load
removal. It is suggested that these types of model will
help the processes involved in disc development and
degeneration to be understood and in the future might
indicate strategies to prevent the degeneration and
possibly to initiate disc repair.
Furlong and Palazotto [78] examined the eVect of
herniation on the compressive characteristics of intervertebral discs of rhesus monkeys. They adjusted the
properties of their axisymmetric viscoelastic nite
element model of healthy and herniated discs until they
agreed with experimental results and then examined the
internal stress distribution predicted by the models.
The eVect of dysfunction of the muscles on the biomechanics of the lumbar spine has been modelled by
Kong et al. [79]. The optimum muscle forces required
for a variety of lifting tasks were modi ed before being
applied to a model of the L3L4 motion segment. The
muscle dysfunction was found to destabilize the spine,
reducing the role of the facet joints and increasing the
load carried by the discs and ligaments as a result.
Sharma et al. [80] examined the role of the ligament
and facets on lumbar spinal stability, in particular with
H03402 IMechE 2002

291

respect to rotational instabilities, spondylolisthesis and

stenosis. Amongst other conclusions, they suggested that
a facet which is more sagittally oriented may be linked
to the cause of spondylolisthesis whereas less transversely oriented joints may be linked to rotational instabilities in extension. Also, facetectomy may restore
spinal canal size without compromising segment stability. In contrast, graded facetectomy on a lumbar segment was considered by Natarajan et al. [81] who
concluded that torsional stability might be jeopardized
if more than 75 per cent of any facet joint was removed.
Yoganandan and colleagues used their model of the
lower cervical spine to investigate the eVects of unilateral
and bilateral graded facetectomy [82] and laminectomy
[83] on the segmental mobility and disc annulus stress.
They found that both the disc stress and the rotational
motion increased signi cantly with bilateral resection of
greater that 50 per cent, with laminectomy also leading
to a marked change in angular motion and disc stress,
particularly in exion. Hirabayashi and Kumano [84]
also modelled double-door laminoplasty where the
spinous processes and laminae are split centrally and
spacers are inserted to keep the laminae apart, thereby
decompressing the nerve tissues. Figure 19 shows the
nite element model used by Hirabayashi and Kumano
and illustrates the position after splitting of the
spinous process.
A non-linear model of L3L5, the sacrum and ilium
was used by Yamamoto et al. [85] to examine spondylolysis in the lower lumbar vertebrae. They concluded
that an extension motion is the most threatening and
that spondylolysis is most likely in L5, which they comment coincides with the clinical observation that the
separation of the interarticular portion often starts at L5.
Finite element models have been used to examine
scoliosis and its treatment. The rst of these were
so-called simple nite element models (because they
used just a few elements to represent each motion segment). For example, Vanderby et al. [86] developed a
two-dimensional model of a patients spine from T5 to
L3 where the properties of each motion segment were
derived from experimental measurements. Viviani et al.
[87] used a similar approach and used the models to
calculate the optimal corrective forces that should be
applied with Harrington distraction rods. After corrective surgery with the rods, the measured corrective deformations were reported to be within 10 per cent of the
nite element predictions. Patwardhan et al. [88] also
examined the performance of two devices for treatment
of scoliosis.
Stokes and Laible [89] and Stokes and GardnerMorse [90] developed several three-dimensional simple
models of the thoracolumbar spine to determine whether
they could simulate the conditions that might lead to
scoliosis, by for example asymmetric growth of the ribs
producing axial rotation and lateral curvature of the
spine. Later, Stokes and Gardner-Morse [91] and
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

292

M J FAGAN, S JULIAN AND A M MOHSEN

Fig. 19

Finite element model of cervical laminoplasty for posterior decompression surgery of the cervical
spine. The gure on the right shows the position of the vertebra after splitting of the spinous process.
(From Hirabayashi and Kumano [84])

Gardner-Morse and Stokes [92] used the same models

to examine both Harrington and CotrelDubousset
instrumentation.
Azegami et al. [93] presented a detailed model of the
whole spine that was used to examine the eVects of
growth (at T4T10 in particular) and the resulting deformation and likely buckling modes of the spine. They
used a simpli ed three-dimensional geometry for each
vertebra but, from the analysis, concluded that the
second bending mode in the coronal plane is one of the
most likely mechanical phenomena in the aetiology of
thoracic idiopathic scoliosis. A simpli ed but functionally representative model of the spine and thorax
was developed by Aubin et al. [94] and Le Borgne [95]
to examine scoliotic deformities and their treatments.
Similarly, a new design of central rod system for treatment of scoliosis was evaluated by Liu et al. [96 ]
and compared with CotrelDubousset instrumentation.
However, because they were only interested in the
performance of the rod systems, their models of the
vertebrae were greatly simpli ed.
Skalli et al. [97] used a detailed model of three vertebrae [29] to examine a CotrelDubousset type of
device and an anterior bone graft applied to stabilize a
fracture of varying severity in the middle vertebra. The
models they used are shown in Fig. 20. A much simpli ed model of the lumbar spine was used by Rohlmann
et al. [98], as illustrated in Fig. 21, to examine the eVects
of a similar xation device on stresses in the vertebral
end plates and intervertebral discs. From their model
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

they concluded that the stiVness of the device had a

signi cant in uence on the bridged discs, but only a
minor in uence on the stresses in adjacent discs. Lim
et al. [99] examined the biomechanics of trans xation
in pedicle screw instrumentation but used a very
idealized model.
Goel et al. [100, 101] used their model of two motion
segments to analyse the use of a screw-plate device based
on the SteVee ( VSP) system and various other fusion
devices. The contact conditions of a posteriorly plated
canine cervical spine (C3C6) were analysed by
Villarraga et al. [102], while Liu et al. [103] used a nite
element analysis to evaluate the biomechanical properties of a new anterior spinal implant for vertebral
fractures in osteoporotic patients.
Polikeit et al. [104] recently presented results of a
detailed study examining the stresses in lumbar vertebrae
after insertion of interbody cages ( Fig. 22). They concluded that the density of the bone underneath the cage
was one of the most important factors determining the
loading and behaviour of the motion segment. A number
of anterior cervical spine interbody fusion devices and
surgical procedures have been analysed by Kumaresan
et al. [105]. They considered ve diVerent types of
interbody fusion material (titanium core, titanium cage,
tricortical iliac crest, tantalum core and tantalum cage)
and two types of surgical procedure. Various loading
conditions were analysed and the stiVness of the spine
was predicted, together with the disc and vertebral
stresses. Posterior lumbar interbody fusion devices were
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

Fig. 20

Three-dimensional models of diVerent fracture conditions used in a study of short spinal xation
devices. (From Skalli et al. [97]).

Fig. 21

A simpli ed model of the lumbar spine with spinal

instrumentation to examine the eVects of the device
on the intervertebral discs. (From Rohlmann et al.
[98])

also analysed by Pitzen et al. [106 ] who considered the

L3L4 motion segment with and without additional posterior instrumentation. Finally, Totoribe et al. [107]
described a model of an L4L5 motion segment which
was modi ed to represent a denucleated disc, posterolateral fusion (with and without denucleation and facet
fusion) and interbody fusion. Their model of posterolateral fusion is shown in Fig. 23 with the bone graft
mass shown separately on the left.
H03402 IMechE 2002

293

CONCLUSIONS

A detailed biomechanical investigation of the human

spine can be undertaken in two ways: by experimentation
or by modelling. The experimental studies can be based
on animal spines, cadaveric spines or arti cial spine
models, while the modelling may be mathematical or
computer based. Of course each approach has a number
of advantages and disadvantages. Experimentation provides the most direct and obvious way to obtain information on spinal biomechanics and performance, but it
can be expensive and unwieldy and suVers from specimen variability and issues of repeatability, although that
variability will itself be an important parameter in the
spines characteristics.
Conversely, modelling has the advantages that an
in nite number and variety of specimens can be generated and examined if required, the same specimen can
be used for any number of loading conditions and combinations, the specimen cannot be overloaded or damaged, and the specimen can be modi ed to re ect any
physiological condition. Modelling can also provide a
wealth of information that experimentation cannot, such
as the stress distribution through the discs and vertebral
bodies. However, generating an accurate model and
ensuring that it is a valid representation of the problem
may be diYcult and, for a problem as complex as the
spine, cannot easily be undertaken. Thus the work that
has been going on over the years towards the establishment of representative spine models is essential if the
models are to be fully understood and truly representative spine models are to be developed. Indeed it seems
that the research has reached the point where the understanding, experience and computing power are now
available, and a number of functionally representative
models have recently appeared.
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

294

M J FAGAN, S JULIAN AND A M MOHSEN

Fig. 22

Fig. 23

A detailed model of lumbar vertebrae with interbody cages. (From Polikeit et al. [104])

Model of posterolateral fusion with the bone graft shown separately ( left). (From Totoribe et al.
[107])

In the development of any nite element model, it is

important to de ne exactly what information is required
from the model and which are the critical parameters
and characteristics for that application and then to
construct the model accordingly.
Finite element models of the spine are complex for a
number of reasons. Firstly, the geometries of the components, and vertebrae in particular, are irregular and,
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

where facet contact between adjacent vertebrae is

important, then these and possibly other regions of the
geometry will need to be modelled accurately. In
addition, the discs have a complex structure and nonlinear behaviour. This non-linearity arises from the nonlinear properties of the constituent materials, especially
the collagen bres, but also the change in orientation
of the bres during loading resulting in non-linear
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

geometry eVects. Non-linear geometry eVects will also

be important when more than one motion segment is
considered and the de ections of the vertebrae will be
signi cant in relation to the overall geometry of the
model. Thus a nite element model of the spine may
have non-linear contact conditions, non-linear material
properties and non-linear geometry. Such a non-linear
model requires an iterative solution and signi cant computing power, which explains why such analyses have
only recently become feasible.
It is a fact that displacements are more accurately
predicted than the stresses and strains in all nite element
models, because the stresses and strains are derived from
and are very sensitive to the relative diVerences in those
displacements. Detailed examination of the convergence
and accuracy of the spine models is very rarely undertaken; instead the models are validated by comparison
with experimentally measured de ections. While this
invariably shows that the overall characteristics of the
model may be reasonable, it does not ensure convergence
and accuracy of the stresses within the model. Indeed,
because of the relatively coarse meshing in the great
majority of current spine models, the absolute values of
stress that they predict should be considered with great
caution; however, where only a qualitative analysis is
required, then they may be adequate. For example, if a
model is modi ed to simulate denucleation of a disc, it
may be enough to demonstrate that the change produces
a doubling (say) in the stress level in the annulus, without
knowing the absolute values.
Validation of any spinal nite element model is essential but extremely diYcult, due to the unavoidable and
often signi cant physiological variation among individuals and specimens. Comparison with experimental data
needs careful interpretation and analysis, because validation with the simplistic loading conditions invariably
applied in a laboratory setting does not necessarily mean
that the model will function correctly with all the complex loading regimes experienced in vivo. Thus it is
important to undertake sensitivity analyses to assess the
eVect of the diVerent parameters in a model and to
identify those parameters that are particularly critical.
If the validity, accuracy and sensitivity of a nite
element model are understood and the model is applied
appropriately, then it can be used with con dence to
produce valuable results.
Finite element models are already making an important contribution to our understanding of the spine and
its components. Many of the investigations use models
that are simpli ed and idealized because of the unparalleled complexity and uncertainty in the geometry, material properties and boundary conditions of these
problems, but this is not peculiar to spinal modelling
problems. Indeed, the idealization is often a strength
when there is such uncertainty and variation between
one individual and another, allowing causeeVect
relationships to be isolated and fully explored, and the
H03402 IMechE 2002

295

inherent variability of experimental tests to be

eliminated.
Finite element modelling provides a wealth of information on our physiological performance, reduces our
dependence on animal and cadaveric experiments and is
an invaluable complement to clinical studies.
It is quite conceivable that it will be possible to generate patient-speci c models in the future that could be
used for patient assessment and even pre- and interoperative planning. More powerful computers and
software will allow more automated generation of nite
element models directly from CT and MRI data using
feature extraction procedures to derive the important
geometrical details. Developments in in vivo non-invasive
characterization of tissues (mainly ligaments and discs)
and facet joints will be required, which can be achieved
by a direct approach or (in part) by an inverse approach
with the help of modelling. Subsequent validation of
individualized models through the use of clinical examination protocols will also need further consideration and
development.
There is no doubt that nite element analysis has
already contributed signi cantly to spine research but
promises to contribute much more in the future.

REFERENCES
1 Brekelmans, W. A. M., Poort, H. W. and SlooV, T. J.
A new method to analyse the mechanical behaviour of
skeletal parts. Acta Orthopaedica Scandinavica, 1972, 43,
301317.
2 Latham, F. A study in body ballistics. Proc. R. Soc. Lond.,
1957, 174, 121.
3 Toth, R. Multiplying degree of freedom, non-linear spinal
model. In Proceedings of the 19th Annual Conference on
Engineering in Medicine and Biology, 1966, p. 8.
4 Orne, D. and Liu, Y. A mathematical model of spinal
response to impact. J. Biomechanics, 1970, 4, 4971.
5 Prasad. P, and King. A. I. An experimentally validated
dynamic model of the human spine. J. Appl. Mechanics,
1974, 546550.
6 Belytschko, T., Schwer, L. and Privitzer, E. Theory and
application of a three-dimensional model of the human
spine. Aviation, Space Environmental Med., 1978, 49,
158165.
7 Williams, J. L. and Belytschko, T. B. A three dimensional
model of the human cervical spine for impact simulation.
J. Biomech. Engng, 1983, 105, 321331.
8 Roberts, S. B. and Chen, P. H. Elastostatic analysis of
the human thoracic skeleton. J. Biomechanics, 1970, 3,
527545.
9 Sundaram, S. H. and Feng, C. C. Finite element analysis
of the human thorax. J. Biomechanics, 1977, 10, 505516.
10 Dietrich, M., Kedzior, K. and Zagrajek, T. A biomechanical model of the human spinal system. Proc. Instn Mech.
Engrs, Part H: J. Engineering in Medicine, 1991, 205,
1926.
11 Dietrich, M., Kedzior, K. and Zagrajek, T. Finite element
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

296

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26
27
28

M J FAGAN, S JULIAN AND A M MOHSEN

method analysis of human spine segment. In International

Series on Biomechanics, Biomechanics XI-A, Proceedings
of the 11th International Congress of Biomechanics, 1987
(Free University Press, Amsterdam).
Lee, M., Kelly, D. W. and Steven, G. P. Model of spine,
ribcage and pelvic responses to a speci c lumbar manipulative force in relaxed subjects. J. Biomechanics, 1995, 28,
14031408.
Hakim, N. S. and King, A. I. A computer aided technique
for generation of a 3-d nite element model of a vertebra.
Computers Biology Med., 1978, 8, 187196.
Hakim, N. S. and King, A. I. A three-dimensional nite
element dynamic response analysis of a vertebra with
experimental veri cation. J. Biomechanics, 1979, 12,
277292.
Yang, K. H., Khalil, T., Tzeng, C. R. and King, A. I. Finite
element model of a functional spinal unit. In Proceedings
of the ASME Biomechanics Seminar, 1983, pp. 137140
(American Society of Mechanical Engineers, New York).
Bozic, K. J., Keyak, J. H., Skinner, H. B., BueV, U. and
Bradford, D. S. Three dimensional nite element modeling
of cervical vertebra: an investigation of burst fracture
mechanism. J. Spinal Disorders, 1994, 7, 102110.
Carter, D. R. and Hayes, W. C. The compressive behaviour of bone as a two phase porous structure. J. Bone Jt
Surg., 1977, 59A, 954962.
Teo, E. C., Paul, J. P. and Evans, J. H. Finite element
analysis of a cadaver second cervical vertebra. Med. Biol.
Engng Computing, 1994, 32, 236238.
Whyne, C. M., Hu, S. S., Kilsch, S. and Lotz, J. C. EVect
of the pedicle and posterior arch on vertebral body
strength predictions in nite element modelling. Spine,
1998, 23, 899907.
Mizrahi, J., Silva, M. J., Keaveny, T. M., Edwards, W. T.
and Hayes, W. C. Finite element stress analysis of normal
and osteoporotic lumbar vertebral body. Spine, 1993, 18,
20882096.
Silva, M. J., Keaveny, T. M. and Hayes, W. C. Load
sharing between the shell and centrum in the lumbar
vertebral body. Spine, 1997, 22, 140150.
Overaker, D. W., Langrana, N. A. and Cuitino, A. M.
Finite element analysis of vertebal body with a non-linear
microstructural model for the trabecular core. J. Biomech.
Engng, 1999, 121, 542550.
Inoue, H., Ohmori, K., Ishida, Y., Suzuki, K., Tanaka, E.
and Murakami, S. Finite element analysis of the lower
lumbar neural arch under facet loading. J. Spinal
Disorders, 1998, 11, 241247.
Goel, V. K., Ramirez, S. A., Kong, W. and Gilbertson,
L. G. Cancellous bone Youngs modulus variation within
the vertebral body of a ligamentous lumbar spineapplication of bone adaptive remodelling concepts. J. Biomed.
Engng, 1995, 117, 266271.
Breau, C., Shirazi-Adl, A. and de Guise, J. Reconstruction
of a human ligamentous lumbar spine using CT images
a three dimensional nite element mesh generation. Ann.
Biomed. Engng, 1991, 19, 291302.
MIMICS (Materialise NV, Leuven).
Fagan, M. J., Julian, S. and Mohsen, A. Unpublished
results.
Stepney, P. D., Acar, B. S., Case, K. and Porter, M.
Computer aided parametric solid modelling of the

Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

29

30
31
32
33
34

35

36
37

38
39
40
41

42
43
44
45

46

spine. In Proceedings of the Eurographics 14th Annual

Conference, 1996, Vol. 2, pp. 235243.
Lavaste, F., Skalli, W., Robin, S., Roy-Camille, R. and
Mazel, C. Three-dimensional geometrical and mechanical
modelling of the lumbar spine. J. Biomechanics, 1992, 25,
11531164.
Robin, S., Skalli, W. and Lavaste, F. In uence of geometrical factors on the behaviour of lumbar spine segments: a
nite element analysis. Eur. Spine J., 1994, 4, 8490.
Belytschko, T., Kulak, R. F. and Schultz, A. B. Finite
element stress analysis of an intervertebral disc.
J. Biomechanics, 1974, 7, 277285.
Kulak, R. F., Belytschko, T. B., Schultz, A. B. and Galante,
J. O. Non-linear behaviour of the human intervertebral
disc under axial load. J. Biomechanics, 1976, 9, 377386.
Lin, H. S., Liu, Y. K., Ray, G. and Nikravesh, P. Systems
identi cation for material properties of the intervertebral
joint. J. Biomechanics, 1978, 11, 114.
Spiker, R. L. A simpli ed hybrid-stress nite element
model of the intervertebral disc. In Finite Elements in
Biomechanics (Eds R. H. Gallagher, B. R. Simon, P. C.
Johnson and J. F. Gross), 1982, pp. 295312.
Spiker, R. L., Jakobs, D. M. and Schultz, A. B. Material
constants for a nite element model of the intervertebral
disc with a bre composite annulus. J. Biomech. Engng,
1986, 108, 111.
Shirazi-Adl, A., Shrivastava, S. C. and Ahmed, A. M.
Stress analysis in the lumbar disc body unit in compression. Spine, 1984, 9, 120134.
Shirazi-Adl, A., Ahmed, A. M. and Shrivastava, S. C. A
nite element study of a lumbar motion segment subjected
to pure sagittal plane moments. J. Biomechanics, 1986,
19, 331350.
Shirazi-Adl, A. Finite element evaluation of contact loads
on facets of an L2L3 lumbar segment in complex loads.
Spine, 1991, 16, 533541.
Shirazi-Adl, A. Load bearing role of facets in a lumbar
segment under sagittal plane loadings. J. Biomechanics,
1987, 20, 601613.
Shirazi-Adl, A. On the bre composite material models of
disc annuluscomparison of predicted stresses.
J. Biomechanics, 1989, 22, 357365.
Shirazi-Adl, A. and Drouin, G. Finite element simulation
of changes in the uid content of human lumbar discs:
mechanical and clinical implications. Spine, 1992, 17,
206212.
Shirazi-Adl, A. Analysis of role of bone compliance on
mechanics of a lumbar motion segment. J. Biomech.
Engng, 1994, 116, 408412.
Kasra, M., Shirazi-Adl, A. and Drouin, G. Dynamics of
human lumbar intervertebral joints experimental and
nite element investigations. Spine, 1991, 17, 93102.
Argoubi, M. and Shirazi-Adl, A. A poroelastic creep
response analysis of a motion segment on compression.
J. Biomechanics, 1996, 29, 13311339.
Rao, A. A. and Dumas, G. A. In uence of material properties on the mechanical behaviour of the L5S1 intervertebral disc in compression: a non-linear nite element
study. J. Biomed. Engng, 1991, 13, 139151.
Simon, B. R., Wu, J. S. S., Carlton, M. W., Evans, J. H.
and Kazarian, L. E. Structural models for human spinal
motion segments based on a poroelastic view of the intervertebral disc. J. Biomech. Engng, 1985, 107, 327335.
H03402 IMechE 2002

FINITE ELEMENT ANALYSIS IN SPINE RESEARCH

47 Laible, J. P., P aster, D. S., Krag, M. H., Simon, B. R.

and Haugh, L. D. A poroelastic nite element model with
application to the intervertebral disc. Spine, 1993, 18,
659670.
48 Duncan, N. A., Ashford, F. A. and Lotz, J. C. Experimental
validation of poroelastic nite element predictions for the
uid pressure and solid stress in the annulus brosis. In
Proceedings of the 1997 Bioengineering Conference, 1997,
BED-Vol. 35, pp. 2526 (American Society of Mechanical
Engineers, New York).
49 Martinez, J. B., Oloyede, V. O. A. and Broom, N. D.
Biomechanics of load bearing of the intervertebral disc:
an experimental and nite element model. Med. Engng
Phys., 1997, 19, 145156.
50 Wang, J. L., Parnianpour, M., Shirazi-Adl, A., Engin,
A. E., Li, S. and Patwardhan, A. Development and validation of a viscoelastic nite element model of an L2/L3
motion segment. Theor. Appl. Fracture Mechanics, 1997,
28, 8193.
51 Wang, J. L., Parnianpour, M., Shirazi-Adl, A. and Engin,
A. E. The dynamic response of L2L3 motion segment in
cyclic axial compressive loading. Clin. Biomechanics, 1998,
13, S16S25.
52 Wang, J. L., Parnianpour, M., Shirazi-Adl, A. and Engin,
A. E. Rate eVect on sharing of passive lumbar motion
segment under load-controlled sagittal exion: viscoelastic
nite element analysis. Theor. Appl. Fracture Mechanics,
1999, 32, 119128.
53 Ueno, K. and Liu, Y. K. A three-dimensional non-linear
nite element model of lumbar intervertibral joint in
torsion. J. Biomech. Engng, 1987, 109, 200209.
54 Lu, Y. M., Hutton, W. C. and Gharpuray, V. M. Do bending, twisting and diurnal uid changes in the disc aVect
the propensity to prolapse? A viscoelastic nite element
model. Spine, 1996, 21, 25702579.
55 Lu, Y. M., Hutton, W. C. and Gharpuray, V. M. The eVect
of uid loss on the viscoelastic behavior of the lumbar
intervertebral disc in compression. J. Biomech. Engng,
1998, 120, 4854.
56 Goel, V. K., Monroe, B. T., Gilbertson, L. G. and
Brinckmann, P. Interlaminar shear stresses and laminae
separation in a disc: nite element analysis of the L3L4
motion segment subjected to axial compressive loads.
Spine, 1995, 20, 689698.
57 Natarajan, R. N. and Andersson, G. B. J. The in uence of
lumbar disc height and cross-sectional area on the
mechanical response of the disc to physiologic loading.
Spine, 1999, 24, 18731881.
58 Natali, A. and Merio, E. Non-linear analysis of an intervertebral disc under dynamic load. J. Biomech. Engng,
1990, 112, 358363.
59 Suwito, W., Keller, T. S., Basu, P. K., Weisberger, A. M.,
Strauss, A. M. and Spengler, D. M. Geometric and material property study of the human lumbar spine using the
nite element method. J. Spinal Disorders, 1992, 5, 5059.
60 Lu, Y. M., Hutton, W. C. and Gharpuray, V. M. Can
variations in intervertebral disc height aVect the mechanical function of the disc? Spine, 1996, 21, 22082216.
61 Goel, V. K., Kong, W., Han, J. S., Weinstein, J. N. and
Gilbertson, L. G. A combined nite element and optimisation investigation of lumbar spinal mechanics with and
without muscles. Spine, 1993, 18, 15311541.
H03402 IMechE 2002

297

62 Kong, W. Z., Goel, V. K., Gilbertson, L. G., Weinstein,

J. N. and Parnianpour, M. EVects of muscle dysfunction
of lumbar spine mechanics: a nite element study based
on a two motion segments model. Spine, 1996, 21,
21972207.
63 Shirazi-Adl, A. and Parnianpour, M. EVect of changes in
lordosis on mechanics of the lumbar spine lumbar curvature in lifting. J. Spinal Disorders, 1999, 12, 436447.
64 Shirazi-Adl, A. and Parnianpour, M. Non-linear response
analysis of the human ligamentous lumbar spine in compression. On mechanisms aVecting the postural stability.
Spine, 1993, 18, 147158.
65 Shirazi-Adl, A. and Parnianpour, M. Stabilising role of
moments and pelvic rotation on the human spine in
compression. J. Biomech. Engng, 1996, 118, 2631.
66 Shirazi-Adl, A. Biomechanics of the lumbar spine in sagittal lateral moments. Spine, 1994, 19, 24072414.
67 Shirazi-Adl, A. Non-linear stress analysis of the whole
lumbar spine in torsionmechanics of facet articulation.
J. Biomechanics, 1994, 27, 289299.
68 Shirazi-Adl, A. and Parnianpour, M. Role of posture in
mechanics of the lumbar spine in compression. J. Spinal
Disorders, 1996, 9, 277286.
69 Kleinberger, M. Application of nite element techniques
to the study of cervical spine mechanics. In Proceedings
of the 37th Stapp Car Crash Conference, Texas, 1993,
pp. 261272.
70 Voo, L. M., Denman, J. A., Kumaresan, S.,
Yoganandan, N., Pintar, F. A. and Cusick, J. F.
Development of 3D nite element model of the cervical
spine. Adv. Bioengng, 1995, 131, 1314.
71 Yoganandan, N., Kumaresan, S., Voo, L. and Pintar, F. A.
Finite element model of the human lower cervical spine:
parametric analysis of the C4C6 unit. J. Biomech. Engng,
1997, 119, 8792.
72 Kumaresan, S., Yoganandan, N. and Pintar, F. A. Finite
element analysis of the cervical spine: a material property
sensitivity analysis. Clin. Biomechanics, 1999, 14, 4153.
73 Maurel, N., Lavaste, F. and Skalli, W. A threedimensional parameterized nite element model of the
lower cervical spine. Study of the in uence of the posterior
articular facets. J. Biomechanics, 1997, 30, 921931.
74 Kumaresan, S., Yoganandan, N. and Pintar, F. A. Finite
element modelling approaches of the human cervical spine
facet joint capsule. J. Biomechanics, 1998, 31, 371376.
75 Goel, V. K. and Clausen, J. D. Prediction of load sharing
among spinal components of a C5C6 motion segment
using the nite element approach. Spine, 1998, 23,
684691.
76 Fagan, M. J., Julian, S. and Mohsen, A. The development
of a computer model to assist in the diagnosis of spinal
conditions. Med. Biol. Engng Computing, 1999, 37,
11781179.
77 Lotz, J. C., Colliou, O. K., Chin, J. R., Duncan, N. A. and
Liebenberg, E. Compression-induced degeneration of the
intervertebral disc: an in vivo mouse model and niteelement study. Spine, 1998, 23, 24932506.
78 Furlong, D. R. and Palazotto, A. N. A nite element analysis of the in uence of surgical herniation on the viscoelastic properties of intervertebral discs. J. Biomechanics,
1983, 16, 785796.
79 Kong, W. Z., Goel, V. K. and Gilbertson, L. G. Prediction
Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

298

80

81

82

83

84

85

86

87

88

89

90

91

92

93

M J FAGAN, S JULIAN AND A M MOHSEN

of biomechanical parameters in the lumbar spine during

sagittal plane lifting. J. Biomech. Engng, 1998, 120,
273280.
Sharma, M., Langrana, N. A. and Rodriguez, J. Role of
ligament and facets in lumbar spinal stability. Spine, 1995,
20, 887900.
Natarajan, R. N., Andersson, G. B. J., Patwardhan, A. G.
and Andriacchi, T. P. Study on eVect of graded facetectomy on change in lumbar motion segment torsional
exibility using three-dimensional continuum contact
representation for facet joints. J. Biomech. Engng, 1999,
121, 215221.
Voo, L. M., Kumaresan, S., Yoganandan, N., Pintar, F. A.
and Cusick, J. F. Finite element analysis of cervical
facetectomy. Spine, 1997, 22, 964969.
Kumaresan, S., Yoganandan, N., Pintar, F. A., Voo, L. M.,
Cusick, J. F. and Larson, S. J. Finite element modelling
of cervical laminectomy with graded facetectomy.
J. Spinal Disorders, 1997, 10, 4046.
Hirabayashi, S. and Kumano, K. Finite element analysis
of the space created by split spinous processes in double
door laminoplasty to optimize shape of an arti cial spacer.
J. Musculoskeletal Res., 2000, 4(1), 4754.
Yamamoto, S., Tanaka, E., Mihara, K., Inoue, H. and
Ohmori, K. Finite element evaluation of spondylolysis
taking account of non-linear mechanical properties of ligaments and annulus brosis. Japan Soc. Mech. Engrs Int.
J., Ser. C, 1999, 42, 521531.
Vanderby, R., Daniele, M., Patwardhan, A. G. and Brunch,
W. H. A method for the identi cation of in vivo segmental
stiVness properties of the spine. J. Biomech. Engng, 1986,
108, 312316.
Viviani, G. R., Ghista, D. N., Lozada, P. J., Subbaraj, K.
and Barnes, G. Biomechanical analysis and simulation of
scoliosis surgical correction. Clin. Orthop. Related Res.,
1986, 208, 4047.
Patwardhan, A. G., Dvonch, V. M., Bunch, W. H.,
Gavin, T., Vanderby, R., Meade, K. P. and Sartori, M. J.
Orthotic stabilization of idiopathic scoliotic curves. In
Adv. Bioengng (Ed. A. G. Erdman), 1987, p. 31.
Stokes, I. A. F. and Laible, J. P. Three dimensional osseoligamentous model of the thorax representing initiation of
scoliosis by asymmetric growth. J. Biomechanics, 1990,
23, 589595.
Stokes, I. A. F. and Gardner-Morse, M. Analysis of the
interaction between vertebral lateral deviation and axial
rotation in scoliosis. J. Biomechanics, 1991, 24, 753759.
Stokes, I. A. F. and Gardner-Morse, M. Three dimensional
simulation of Harrington distraction instrumentation
for surgical correction of scoliosis. Spine, 1993, 18,
24572464.
Gardner-Morse, M. and Stokes, I. A. F. Three dimensional
simulations of the scoliosis derotation manoeuvre with
CotrelDubousset instrumentation. J. Biomechanics,
1994, 27, 177181.
Azegami, H., Murachi, S., Kitoh, J., Ishida, Y.,
Kawakami, N. and Makimo, M. Etiology of idiopathic

Proc Instn Mech Engrs Vol 216 Part H: J Engineering in Medicine

94

95

96

97

98

99
100
101

102

103
104

105
106

107

scoliosis computational study. Clin. Orthop. Related

Res., 1998, 357, 229236.
Aubin, C. E., Descrimes, J. L., Dansereau, J., Skalli, W.,
Lavaste, F. and Labelle, H. Modelisation geometrique
du rachis et du thorax pour lanalyse biomecanique
par elements nis des deformations scoliotiques. Ann.
Chirurgie, 1995, 49, 749761.
Le Borgne, P., Skalli, W., Dubousset, J., Dansereau, J.,
Zeller, R. and Lavaste, F. Finite element model of scoliotic
spine: mechanical personalization. In Research into
Spinal Deformities, Vol. 2 (Ed. I. A. F. Stokes), 1999,
pp. 130133.
Liu, C. L., Kao, H. C., Wang, S. T., Lo, W. H. and Cheng,
C. K. Biomechanical evaluation of a central rod system in
the treatment of scoliosis. Clin. Biomechanics, 1998, 13,
548559.
Skalli, W., Robin, S., Lavaste, F. and Dubousset, J. A
biomechanical analysis of short segment spinal xation
using a three dimensional geometric and mechanical
model. Spine, 1993, 18, 536545.
Rohlmann, A., Calisse, J., Bergmann, G. and Weber, U.
Internal spinal xator stiVness has only a minor in uence
on stresses in the adjacent discs. Spine, 1999, 24,
11921196.
Lim, T. H., Eck, J. C., An, H. S., Hong, J. H., Ahn, J. Y.
and You, J. W. Biomechanics of trans xation in pedicle
screw instrumentation. Spine, 1996, 21, 22242229.
Goel, V. K., Lim, T. H., Gwon, J., Chen, J. Y. and Han, J.
Biomechanics of fusion. In Spinal Stenosis (Eds. G. B.
Andersson and T. W. McNeill ), 1992, pp. 403414.
Goel, V. K., Lim, T. H., Gwon, J., Chen, J. Y. and Han, J.
EVects of rigidity of an internal xation devicea comprehensive biomechanical investigation. Spine, 1991, 16,
155S161S.
Villarraga, M. L., Anderson, R. C., Hart, R. T. and Dinh,
D. H. Contact analysis of a posterior cervical spine plate
using a three-dimensional canine nite element model.
J. Biomech. Engng, 1999, 121, 206214.
Liu, C. L., Chen, H. H., Cheng, C. K., Kao, H. C. and Lo,
W. H. Biomechanical evaluation of a new anterior spinal
implant. Clin. Biomechanics, 1998, 13, S40S45.
Polikeit, A., Orr, T. E. and Nolte, L. P. Factors in uencing
the stresses in the lumbar vertebra after insertion of
interbody cages: nite element analyses. In Proceedings
of the 12th Conference of the European Society of
Biomechanics, Dublin, 2000, p. 168.
Kumaresan, S., Yoganandan, N. and Pintar, F. A. Finite
element analysis of anterior cervical spine interbody
fusion. Biomed. Mater. Engng, 1997, 7, 221230.
Pitzen, T., Matthis, D., Muller-Storz, H., Caspar, W.,
Steudel, W. I. and Harms, J. Die Primarstabilitat von zwei
PLIF-TechnikenEine biomechanische und Finite
Element Analyse. Zentbl. Neurochirurg., 1999, 60,
114120.
Totoribe, K., Tajima, N. and Chosa, E. A biomechanical
study of posterolateral lumbar fusion using a threedimensional nonlinear nite element method. J. Orthop.
Sci., 1999, 4, 15126.

H03402 IMechE 2002