Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Submitted by
Abel Zhou
Faculty of Health
The University of Canberra, Canberra, Australia
A sample grid with a cut
i
Contents
ACKNOWLEDGEMENTS ----------------------------------------------------------------------------- XV
v
1.2.7. Anti-scatter grids -------------------------------------------------------------------------------- 21
1.2.7.1. Mechanism of scatter radiation reduction by grids ----------------------------------------------------------- 21
1.2.7.2. Reduction of primary and scatter radiation by grids --------------------------------------------------------- 23
1.2.7.3. Patient radiation dose increase attributed to grids ------------------------------------------------------------ 24
1.5. PURPOSE------------------------------------------------------------------------------------------------ 25
vi
2.4.3. Broad beam condition -------------------------------------------------------------------------- 43
SUMMARY --------------------------------------------------------------------------------------------------- 45
vii
3.5.4. KERMA (kinetic energy released per unit mass) ------------------------------------------- 72
3.5.5. Absorbed dose ----------------------------------------------------------------------------------- 73
3.5.6. Radiation exposure------------------------------------------------------------------------------ 74
SUMMARY --------------------------------------------------------------------------------------------------- 76
GRIDS --------------------------------------------------------------------------------------------------------- 79
4.2. NEW METHOD FOR DETERMINING RADIATION TRANSMISSION IN GRID DESIGNS -------------- 82
4.3.1. Criteria for comparisons of grid performance and grid selection ----------------------- 93
4.3.2. Grid details, phantoms, radiation beam quality, and simulation setups ---------------- 93
4.3.3. Simulation of grid movement ------------------------------------------------------------------ 94
4.3.4. Simulation of photon transport ---------------------------------------------------------------- 96
4.3.5. Results -------------------------------------------------------------------------------------------- 96
4.3.5.1. Theoretical comparison: transmission in parallel grid ------------------------------------------------------- 96
4.3.5.2. Transmission of primary radiation in focused grid: Tp------------------------------------------------------ 99
4.3.5.3. Transmission of scatter radiation in focused grid: Ts ------------------------------------------------------ 100
4.3.5.4. Images behind a stationary grid and moving grid ---------------------------------------------------------- 103
4.3.5.5. Dependence of radiation transmission (Tp and Ts) -------------------------------------------------------- 104
4.3.6. Comparisons of Tp and Ts between simulation and the literature --------------------- 107
4.3.6.1. Tp comparison---------------------------------------------------------------------------------------------------- 107
4.3.6.1. Ts comparison ---------------------------------------------------------------------------------------------------- 109
viii
SUMMARY ------------------------------------------------------------------------------------------------- 111
ix
CHAPTER 7: MONTE CARLO EVALUATION OF NEW GRID DESIGNS AND
x
7.3.1.2. New mammographic grid designs vs the literature: K SNR ------------------------------------------------- 167
xi
Acknowledgements
Exploring new things or discovering new knowledge is a relentless journey. Four years ago, when
I was doing my master’s at Charles Sturt University, Professor Rob Davidson inspired me to
undertake a research journey. I would like to thank Professor Davidson for his encouragement and
inspiration. Since then, I have received continuous support from many people.
Professor Rob Davidson has given me his endless support since the start of my research journey.
This doctoral project continued from an original master’s project which began from an interest
arising from my eight years’ experiences in medical x-ray imaging as a radiographer and later as
a clinical application specialist for the Dextroscope, a virtual reality system for neurosurgical
planning. Professor Davidson’s expertise in medical x-ray imaging, his advice on my research
progress, and his vision and direction for research inspired me to accomplish this doctoral project.
I would like to thank Professor Davidson for the support in allowing me to use the radiation science
labs both at Charles Sturt University and the University of Canberra, for organising experiments;
data collection at The Canberra Hospital, for a scholarship from the Faculty of Health at the
University of Canberra (without this scholarship, I could not have completed this project), for the
inspiration in developing the methods used in this project, for his friendship in supporting me
throughout my candidature, and for other support and help not yet mentioned.
Special thanks goes to Mr Yuming Yin, who helped me with the radiation transmission method
used in my Monte Carlo simulation. I would like to thank Dr Graeme L White for the comments
on and suggestions for my articles and thesis and for support in performing the experiments at The
Canberra Hospital. I would like to thank Mr Bill Shelley and his colleagues from the Faculty of
Arts and Design at the University of Canberra for helping make the radiation blockers. I would
xv
like to thank the friendly staff from the Department of Radiology at The Canberra Hospital for
I would like to thank my family for supporting me in undertaking a journey in research. Dora and
my children, Samantha and Albert, have understood my desire to explore new things and have
allowed me time that could have been spent with them. Mum and Dad are always encouraging me
exploring and learning. Sorry Mum, I should have spent more time with you in China during your
Finally, I would like to thank my supervisors once more for spending so much of their time to help
me. At Charles Sturt University, thank you, Dr Han Swan, for your help with statistics. Thank you,
Dr Kelly Spuur, for your expert help in mammography. Mr Warren Lusby, thank you for your
expert help in anatomy and radiographic positioning. Dr Allan Ernest, Dr Xiaocheng Zhu, Dr
Lihong Zheng, Dr De Li Liu, and Dr Liang Bin, thank you for your friendships. Dr Liang Bin,
thank you for helping me with MatLab. Following my transfer of candidature to the University of
Canberra, I would like to thank Associate Professor Stuart Semple, my supervisor, and Dr Richard
Keegan for supporting my candidature. Thank you, Dr Peter Copeman, for your expert help in
xvi
Abstract
Medical imaging is commonly used to provide clinicians with a diagnosis for their patients, or to
follow up on a patient’s treatment. Medical imaging incorporates the use of x-rays, ionising
radiation, and non-ionising radiation. One of the major challenges in x-ray imaging is to
substantially remove scatter radiation reaching the image receptor without increasing a patient’s
primary radiation and the matter of the imaging anatomy. These interactions may scatter primary
radiation and produce scatter radiation, some of which moves towards the image receptor.
Attenuated primary radiation and scatter radiation reaching the image receptor form the image.
The intensities of the scatter radiation deform the differentiation attenuations in the attenuated
primary radiation. Consequently, the image diagnostic quality is reduced. A common solution for
removing scatter radiation reaching the image receptor is to use an anti-scatter grid technique. This
This project investigated how a grid technique can be used in mammography and general
radiography without this increase in radiation, and without sacrificing image diagnostic quality.
The project addressed the transmission of primary radiation (Tp) and the transmission of scatter
radiation (Ts) in grid materials. In addition, it presented a quantitative relationship between these
transmissions and the stochastic effects of ionising radiation. This relationship suggests a strategy
for the reduction of these effects, an essential element for patient radiation protection outlined in
xvii
the ALARA (as low as reasonably achievable) principle. This project emanates from the scholarly
1) Using grids fails to minimise radiation exposure delivered to patients because of the
compensation for the reduction of primary radiation in grid materials.
2) Using grids lacks adequate optimisation of image diagnostic quality because of the
remaining scatter radiation in the image.
In this project, a series of actions to investigate grids were performed in three distinct phases. The
first phase examined quantitative methodologies determining Tp and Ts. Moreover, this phase
developed a new Monte Carlo simulation code system. In addition, this phase proposed a novel
criterion for the determination of strip optimal thicknesses by using the first differential of the
quantum signal-to-noise ratio (SNR) improvement factor (KSNR). Furthermore, this phase also
limitations in current radiation transmission methods. An article related to the development and
The second phase of this project developed a new grid design method that can be used to design
grids to overcome the reduction of primary radiation in grid strips. This phase first analysed the
relationships between transmitting primary radiation and reducing scatter radiation. Factors
associated with these relationships were then revealed and issues for grid design were examined.
Finally, the criteria for designing new grids were evaluated by analysing the relationships between
these factors.
xviii
The third phase of this project determined a solution for reducing scatter radiation reaching the
image receptor. This solution neither increases patient exposure to radiation nor sacrifices image
diagnostic quality. In this phase, many new mammographic grids and general grids were designed
using the criteria identified in the second phase of the project. The strip optimal thickness of these
new grids was determined using the strip optimal thickness criterion proposed in the first phase.
The designs of these new grids were evaluated using the Monte Carlo simulation code system
developed and validated in the first phase. The results of these new grids were presented in terms
The results for these new grid designs showed that it is possible to have Tp approximately equal
to the Tp of a perfect grid (Tp = 1). The Ts of these new designs depends on their grid ratio: the
higher the grid ratio, the smaller the Ts. Compared to contemporary grids in the literature, new
general grid designs and new mammographic grid designs have 39% and 28% greater Tp,
respectively, and all of them have smaller Ts. Furthermore, the KSNR of these new designs showed
that new grids could have a benefit of improved image diagnostic quality, regardless of anatomical
mammographic grids have improved image diagnostic quality only for thick breasts (greater than
approximately 5 cm). The new mammographic grid designs, however, will have improved image
diagnostic quality for all breasts, no matter whether they are thin or thick.
In conclusion, the new grid designs proposed in this project substantially remove scatter radiation
reaching the image receptor, do not increase radiation exposure to patients, and do not compromise
image diagnostic quality. Using new grids that use these new designs will not increase the
stochastic effects of ionising radiation in either mammography or general radiography. Using such
xix
new grids to replace current contemporary grids will not only result in less scatter radiation
reaching the x-ray image receptor, but the stochastic effects will also be reduced by more than
radiation reaching the x-ray image receptor can be further reduced by using such new grids with a
Using grids built with these new designs, an effective ionising radiation reduction management
strategy in the implementation of the ALARA principle can be achieved. This means that the
lowest radiation exposure can be achieved using such new grids to produce x-ray images without
xx
Table of Abbreviations
ALARA As low as reasonably achievable
CT Computed tomography
LNT Linear-no-threshold
xxi
Glossary of terms
absorbed dose the energy deposited through ionisation in unit mass of irradiated
material
air-gap technique a way of reducing scatter radiation reaching the image receptor by
keeping a gap in between the imaging object and the image receptor
ALARA (as low as a radiation protection principle, which requires x-ray examinations
reasonably to be performed with the lowest possible radiation dose delivered to
achievable) principle the patient and without compromising the image diagnostic quality
anti-scatter grid a device used in x-ray imaging systems to reduce scatter radiation
reaching the image receptor
area density also known as density thickness, it is calculated as the mass per unit
area. The SI derived unit is kilogram per square metre (kg.m-2)
azimuth angle (vector an angle, in a 3-dimensional Cartesian system, with its z-axis
in 3-dimensional alinged in the vector’s direction and the tail of the vector fixed at
Cartesian system) the Cartesion system’s origin, between the positive x-axis and the
perpendicular projecton of the vector down onto the x-y plane
xxiii
cancerous risk a chance of developing cancer
collimator a device that is placed between the x-ray tube and the patient and
which uses high attenuation materials, typically lead, to control the
shape and size of the x-ray field
computed tomography an x-ray imaging system that uses x-rays to create cross-section
images of the body
digital imaging system an x-ray imaging system that captures and displays images by
digital means. A digital radiographic system is an example of a
digital imaging system
direction cosines (of a the cosines of the angles between the vector and the three
vector) coordinate axes
xxiv
DNA a molecule that carries the genetic instructions used in the growth,
(deoxyribonucleic development, functioning, and reproduction of all known living
acid) organisms and many viruses. It has four bases – adenine (A),
cytosine (C), guanine (G), and thymine (T), which attach to a sugar-
phosphate to form a complete nucleotide
electromagnetic (EM) a type of radiation that consists of electric and magnetic fields at 90
radiation degrees to each other. x-rays, light, and radio waves are examples of
EM radiation
energy fluence the amount of energy passing through a unit cross-section area
exposure time the period during which x-rays are emitted from the x-ray tube
filtration the removal of x-ray photons from the beam by attenuation when
the beam is passed through a medium
focal spot the area on the anode of the x-ray tube where x-ray photons are
produced
free path length (x- a distance that the x-ray passes without interactions
ray)
xxv
image clarity (x-ray) the quality of being easy to visualise anatomic structures in the x-
ray image
image contrast (digital the difference between the intensities of adjacent regions in the
image) image
image diagnostic the image quality that is formed by only the intensities of the
quality attenuated priamry radiation behind the imaging anatomy
latent image the period between exposure to ionising radiation and the
appearance of diseases, such as cancer
xxvi
mass energy the fraction of mass attenuation coefficient that gives rise to the
absorption coefficient total energy locally deposited at the site of interaction
mass energy transfer the fraction of mass attenuation coefficient that gives rise to the
coefficient kinetic energy of the charged particle, and, if any, the charged
particle binding energy
mono-energetic beam an x-ray beam comprising photons of all the same energy
strip optimal thickness for a given grid ratio and strip height, the grid strip thickness with
which the grid has the highest value of the quantum-to-signal-noise
improvement factor
photoelectric effect the emission of electrons when photons are absorbed in matter
(photoelectric
absorption)
xxvii
photon fluence the number of photons or particles passing through a unit cross-
section area
Poisson mottle the appearance of noise in an x-ray image because of low signal-to-
noise ratio in the x-ray intensities reaching the image receptor
polar angle (vector in the photon’s angle in a 3-dimensional Cartesian system that is
3-dimensional system) formed between the photon’s direction and the z-axis positive
direction
primary radiation radiation that does not change direction during its passage
radiation exposure a measure of the ionisation of air ascribed to ionising radiation from
gamma rays or x-rays
xxviii
radiographer a qualified person who is responsible for undertaking a radiographic
examination
Rayleigh scattering the scattering of photons by atoms, without changing the photon
energy
scatter-to-primary the ratio of the amount of scatter radiation to the amount of primary
ratio (SPR) radiation
signal-to-noise ratio a ratio of the amount of signal to the amount of noise in a system
(SNR)
slot technique a way of reducing scatter radiation reaching the image receptor by
constraining the x-ray beam in a narrow fan shape
stochastic effect health risk effect that occurs by chance after a latent period and
without a threshold dose, and whose probability is proportional to
the cumulative dose
xxix
subject contrast the difference in exit intensities between adjacent areas in the x-ray
beam that results from different properties within the anatomy
transmission of the ratio of the amount of primary radiation reaching the image
primary radiation (Tp) receptor with a grid to without a grid
transmission of scatter the ratio of the amount of scatter radiation reaching the image
radiation (Ts) receptor with a grid to without a grid
tube heat loading (x- energy deposition on the anode of the x-ray tube
ray)
tube voltage (kVp) the voltage or electrical potential difference that is applied between
the cathode and the anode of the x-ray tube to accelerate the
electrons between the cathode and the anode to generate x-ray
radiation, usually measured in peak kilovolts (kVp)
x-ray tube a device that generates an x-ray beam and requires a high voltage, a
current, a means of first accelerating then decelerating the electrons,
and a vacuum to achieve the production of x-rays
xxx
Chapter 1: The nature and scope of this project
1.1. Introduction
A significant challenge in x-ray imaging is to optimise radiation protection, taking into account
economic and social factors, using the guiding principle of ‘as low as reasonably achievable’
principle requires delivering the lowest radiation dose to patients when performing x-ray
This project addresses the problem of an anti-scatter grid technique that is commonly used in many
x-ray imaging systems such as mammography and radiography. While this technique reduces
scatter radiation reaching the image receptor, it also reduces primary radiation reaching the image
receptor. Compensation for the reduction in primary radiation is needed to avoid increasing the
image quantum noise, but this compensation increases patients’ radiation exposure.
1.2. Background
1.2.1. Overview of medical x-ray imaging
X-rays were discovered by Röntgen (1896), a German physicist and mechanical engineer. He used
the mathematical designation ‘x’ for something unknown when he speculated on a new kind of ray
from electrical discharge passing through a vacuum tube. These new rays caused shimmering in
the barium platinocyanide screen. Röntgen used x-rays to produce an x-ray image of his wife’s
1
hand, and this image gave birth to a new tool used in the field of medical diagnostics. In 1901,
Since the last century, x-rays have been used for the diagnosis and treatment of many conditions.
In the area of medical diagnostics, x-rays1 are used in a wide variety of imaging technologies, for
example, mammography, to produce diagnostic images. Such technologies have been described in
detail in several textbooks on medical imaging physics (Bushberg et al., 2011, Flower, 2012,
Bushong, 2013). Some x-ray diagnostic imaging technologies are given in Table 1.1.
Table 1.1. Types of medical diagnostic x-ray imaging technologies (Bushberg et al.,
2011).
Imaging technology Characteristics of images
basic components are the x-ray tube, imaging anatomy, and image receptor. First, primary
radiation is generated at the x-ray focal spot. Then the primary radiation is constrained using high
atomic material, such as lead diaphragms, in a cone beam, rectangle beam, or another beam shape.
1
In this thesis, the terms ‘X-rays’, ‘photons’, and ‘radiation’ are used interchangeably, unless otherwise stated.
2
When the primary radiation is travelling in the anatomy, some primary radiation in the beam
interacts with the anatomy’s matter. Differentiation attenuations emanate from the interactions
between the primary radiation and the anatomy’s matter. These interactions may scatter the
primary radiation and produce scatter radiation, some of which moves towards the image receptor.
Attenuated primary radiation and scatter radiation reaching the image receptor form the image.
In x-ray imaging, photon energy is often measured in kiloelectron volts (keV). Photons are
generated from the x-ray tube, and their energy is approximately 10–150 keV. These photons may
interact with tissue through Rayleigh scattering, Compton scattering, or photoelectric absorption
Atoms can be approximately described using the Rutherford-Bohr atomic model (Podgoršak,
2006). In this model, an electron is bound to the nucleus of an atom. The electron is orbiting the
nucleus in one of the atom’s shells. The electron’s orbiting force is the attraction force between
the positive charge of protons in the nucleus and the negative charge of the electron.
atom and causes the electron to be ejected from the atom. The photon’s ability to eject bound
electrons is known as the ionising effect. Photons that can cause the ionising effect are known as
Rayleigh scattering is known as coherent interaction in that an incident photon is deflected and
this photon’s energy is preserved. As illustrated in Figure 1.1, in Rayleigh scattering, the electrons
3
of an atom act as a whole to interact with an incident photon and change the photon’s direction.
During this interaction, no electrons are ejected, and atom ionisation does not occur (Bushberg et
Figure 1.1. Rayleigh scattering. The electrons of an atom act as a whole to interact with an
incident photon and change this photon’s direction.
Rayleigh scattering dominates mainly in x-ray photons with very low energy, such as photon
energy of 10–35 keV used in mammography. In soft tissue, Rayleigh scattering accounts for less
than 5% of interactions for photon energy greater than 70 keV and for about 10% for photon energy
Compton scattering is an incoherent interaction in that a photon interacts with an electron and,
after the interaction, this photon’s direction and energy are changed, as illustrated in Figure 1.2. In
Compton scattering, a photon is more likely to interact with a loosely bound or ‘free’ electron.
4
Figure 1.2. Compton scattering. An incident photon of energy E0 interacts with a bound
electron of binding energy Eb. After the interaction, the scattered photon with energy E c is
deflected away from this incident photon’s direction. The ejected electron is emitted with kinetic
energy Ee-. The direction of this electron forms an angle (β) with this incident photon’s direction.
β is less than 90 degrees.
During Compton scattering, part of the incident photon’s energy is transferred to the bound
electron (Bushberg et al., 2011, Carlton et al., 2012). This electron gains kinetic energy and escapes
from the atom. The kinetic energy plus this electron’s binding energy equals the energy lost by the
photon. The photon is scattered and becomes a scatter photon. This electron’s direction is confined
to an angle, which is not more than π/2 radians with respect to this photon’s direction before the
interaction (Attix, 2008). Compton scattering results in scatter radiation and ionisation.
interacts with a bound electron. The electron absorbs all the photon’s energy and then escapes from
the atom. The electron’s kinetic energy equals the photon’s energy minus the electron’s binding
5
energy (Figure 1.3). Photoelectric absorption only occurs if the photon energy is greater than the
In photoelectric absorption, a photon most likely interacts with the electrons whose binding
energies are the closest to, but less than, the photon’s energy. As illustrated in Figure 1.3, photons
with energy exceeding the K-shell binding energy are most likely to interact with the K-shell
Following photoelectric absorption with a K-shell electron, the atom is ionised with a K-shell
electron vacancy. This vacancy will be most likely filled with an electron from a nearby shell (in
this example, the L-shell) with a lower binding energy. The electron filling this vacancy, however,
can be from any shell (M-shell or N-shell) with a lower binding energy. As it escapes from its shell,
this electron creates another vacancy, which then is filled with an electron from an even lower
binding energy shell. Thus, an electron cascade from outer shells to inner shells occurs.
characteristic x-rays. A bound electron, typically from the same shell of the cascading electron,
may absorb the characteristic x-ray that is emitted by the cascading electron. If, after absorbing the
characteristic x-ray, this bound electron escapes from the shell, this electron then is known as an
6
Figure 1.3. Photoelectric absorption. A 100-keV photon is undergoing an interaction with an
atom of 4-electron shell. In this example, a K-shell electron is ejected with a kinetic energy
equal to the incident photon’s energy minus the K-shell binding energy. The vacancy created
in the K-shell results in the electron’s transition from the L-shell to the K-shell. The difference
in the K-shell and L-shell binding energies results in the emission of a 28 keV characteristic x-
ray. The electron cascade continues and results in the production of other characteristic x-rays
of lower energies; for example, 4 keV and 1 keV. Auger electrons of various energies may be
emitted in lieu of the characteristic x-ray emissions (not shown in this diagram).
The emission of Auger electrons and the emission of characteristic x-rays are competing processes.
The probability of characteristic x-ray emission decreases as the atomic number of the interaction
material decreases. Soft tissue is mostly composed of materials of low atomic matter.
Characteristic x-ray emission does not frequently occur in soft tissue; Auger electron emission
Photons may interact with anatomical matter when they travel, which may cause ionisation in the
anatomy’s tissues or organs. This ionisation has detrimental effects, for example, breaking
deoxyribonucleic acid (DNA) strands or killing cells (Graham et al., 2003, Bushberg et al., 2011),
7
which can be categorised as deterministic effects or stochastic effects (International Commission
Deterministic effects of ionising radiation are acute damage to organs and tissues. This damage
occurs as loss of tissue function or organ function, such as cell death and, in extreme cases, death
of the irradiated individual. Deterministic effects have a threshold dose, and the estimated
threshold doses for various tissues and organs can be found in the International Commission on
The threshold doses of the deterministic effects depend on several factors, including the type of
the ionising radiation, the type of irradiated organ/tissue, the volume of the tissue exposed, and the
clinical effect on the organ/tissue exposed. When the patient’s radiation dose exceeds the threshold
dose, the severity of the deterministic effect increases as the patient’s radiation dose increases
(Khong et al., 2013, International Commission on Radiation Protection, 2007b, Hall, 2006).
The estimates of threshold doses may be lower than what is recommended. Investigations have
shown that radiation-induced eye cataracts and circulatory disease occur at lower radiation doses
than their current threshold doses (International Commission on Radiological Protection, 2007b).
Stochastic effects of ionising radiation are health risks, including cancerous risk, heritable risk,
and non-cancerous risk (Kadhim et al., 2013). These risks usually have a latent period, which may
8
be from several years to more than ten years (International Commission on Radiation Protection,
2007b).
The cancerous risk results from damage to genes by direct or indirect radiation energy deposited
in the nuclear DNA. Unrepaired or wrongly repaired damage in a cell’s DNA may result in the
development of cancer in the individual exposed (Kadhim et al., 2013), known as a somatic effect.
If the damage results in any disease in future generations of the individual exposed, this damage
heritable risk of radiation exposure is mostly evidenced by the offspring of Japanese atomic-bomb
survivors (Preston et al., 1994, Preston et al., 2003, Preston et al., 2007). The non-cancerous risk
responses, and myocardial infarctions (Schultz-Hector and Trott, 2007, Little et al., 2008b, Little,
The probability of the occurrence of stochastic effects is proportionally related to the individual’s
cumulative radiation dose. The severity of the stochastic effects is not related to radiation dose.
For example, a cancer induced by 2 Gray (Gy) is no worse than a cancer induced by 0.01 Gy.
Stochastic effects have no threshold dose, which means a single instance of unrepaired damage in
a cell’s genetic material could cause cancer or a hereditary defect (Mossman, 2006, Tubiana et al.,
2006).
The stochastic effects of patient radiation exposure are evidenced by a wide range of investigations,
such as cancer incidence in the offspring of the Japanese atomic-bomb survivors (Little et al.,
9
2009b, Little, 2003), development of cancers from experimental animals (International
Commission on Radiological Protection, 2007b), and the significantly high rates of cancer among
irradiated populations. These irradiated populations include patients who have undergone long-
term fluoroscopy studies, been exposed to radiation at an early age, or been treated with radiation
therapy, as well as individuals working near accelerators and other sources of ionising radiation.
Among female tuberculosis patients who have undergone extensive diagnostic fluoroscopy, an
increase in breast cancer has been reported in a 30-year follow-up investigation of 4940 women
(Boice Jr et al., 1991). This investigation showed significantly higher rates of breast cancer (147
breast cancers occurred in contrast to the 113.6 expected), which was not apparent until about 10–
15 years after the initial fluoroscopy examinations. No higher rates of breast cancer (87 observed
versus 100.9 expected) were found among women treated without radiation exposure. The results
also showed that age at radiation exposure strongly influenced the risk of radiation-induced breast
cancer, with young women being at the highest risk and those over the age of 40 being at lowest
risk. This finding is supported by a United States scoliosis cohort study of 5573 females who were
exposed to radiation of a much lower cumulative dose before the age of 20 and had a statistically
significant increased risk of breast cancer (Doody et al., 2000). The estimated mean-cumulative
radiation dose to the breast was 108 mGy. When the scoliosis cohort study was limited to
individuals with breast doses between 10 mGy and 90 mGy, the significantly higher risk remained.
Higher cancer risks were reported for various types of cancers in patients who had undergone
radiation therapy. After treatment of haemangioma, a significantly higher risk for breast cancer
was found for a mean dose of 290 mGy to the breast with a range of 0–35.8 Gy at 95% confidence
interval (Eidemüller et al., 2009, Eidemüller et al., 2011). Similar results were found in females
10
who were treated for postpartum mastitis with doses typically from 1 to 6 Gy (Hall, 2006). An
increase in lung cancer had been reported in patients with Hodgkin’s lymphoma who had received
treatment by irradiation doses of 5 Gy or more (Travis et al., 2002, Dores et al., 2002).
Leukaemia is one of the malignant cancers most likely to be linked to radiation exposure. It is
commonly diagnosed in x-ray workers, physicists, and engineers working near accelerators and
other sources of ionising radiation (Little et al., 2009a, Hall and Giaccia, 2012).
The risk of cancer due to ionising radiation exposure has been widely reported and supported with
evidence. The stochastic effects of radiation exposure are unavoidable. One of the essential
The linear-no-threshold (LNT) theory predicts that stochastic effects are proportional to the
cumulative radiation dose. A simple causative relationship between radiation dose and cancer risk
is found from most reliable animal data, excluding mouse data for thymic lymphoma and ovarian
The International Commission on Radiation Protection (2007a) uses dose and dose rate
effectiveness factor (DDREF) values to model the risks of radiation-induced cancer. DDREF
values for doses at or below 2 Gy have a value of 2. In comparison, the Biological Effects of
Ionisation Radiation (BEIR) Committee uses the Bayesian statistics of the combination of the life
span study of atomic-bomb survivors and selected animal studies (National Research Council,
11
2006). The BEIR VII study (National Research Council, 2006) claimed that DDREF values in the
the BEIR VII (National Research Council, 2006) uses a DDREF value of 1.5 to predict cancer risk
radiological protection. The predictions of sex-averaged nominal risks of the stochastic effects of
The stochastic effects of patient radiation exposures are unavoidable and are proportional to the
patient’s cumulative radiation dose. The International Commission on Radiation Protection (2007a)
has developed the radiation protection guidance for the practice of medical x-ray imaging.
The prime goal of radiation protection is to prevent the occurrence of deterministic effects and to
minimise the stochastic effects, not only for patients but also for healthcare professionals. The
medical x-ray examinations that relate to justification, optimisation, and dose minimisation. The
justification refers to the fact that every radiation exposure that a patient receives must be
associated with a positive net benefit. The optimisation and dose minimisation mean that all
radiation exposures must be kept ALARA without compromising image diagnostic quality.
12
A significant challenge with the ALARA principle is to produce the acceptable image diagnostic
image. Image contrast is the difference between adjacent intensities in an image. Adequate contrast
can increase clinicians’ confidence in making diagnoses. In the process of image production,
contrast can be described at each production stage. In x-ray imaging, two common concepts of
Subject contrast is the fundamental contrast that emanates from the intensities of the x-ray beam
exiting the imaging object. Subject contrast is converted to a latent image stored in the image
receptor, and a final image is processed from the latent image. The difference observed in the final
image is the image contrast. In film-screen systems, the image contrast processed from the latent
image cannot be altered. In digital systems, however, the image contrast can be manipulated when
as the difference in intensities between adjacent areas in the exit beam. Factors influencing the
subject contrast are photon energy and the tissues. A tissue’s physical properties, including the
A tissue’s physical properties generally cannot be altered. Compressions can be applied in limited
x-ray examinations to enhance the subject contrast by changing the tissue thickness; for example,
mammography with a compression technique (OLeary et al., 2011). Subject contrast can be
13
enhanced by altering the physical properties by using a contrast agent, such as potassium iodine-
sulphate or barium solutions. A natural way of enhancing the subject contrast is introducing air;
Photon energy can be altered to enhance subject contrast because the probability of interactions
between photons and tissue depends on the photons’ energy. An x-ray beam generated with a peak
tube voltage has a range of photon energies. This beam is a poly-energetic beam and is
characterised by the anode materials, filtration materials, and peak kilovoltage (kVp) applied to
the x-ray generator. The beam’s properties can be measured in a spectrum. Most x-ray generators
are built with fixed anode materials and filtration materials, but some may be constructed with
represents a two-dimensional image as a number of rows and columns of pixels. For digital x-ray
images, the image contrast is the difference between the pixel values of adjacent regions in the
A digital image’s pixel values can be altered by changing the window level and width of the
image’s display. A digital image window is the range of values for the pixels. A digital image
window’s level is the middle point of this range, and a digital image window’s width is the width
of this range.
Digital images’ pixel values can be changed by advanced image processing. There are many
advanced image processing algorithms, such as look-up tables, histograms, spatial enhancement,
14
unsharp masks, contrast-enhancement masks, dynamic range control, and multi-scale processing
(Davidson, 2006).
Radiographic contrast of x-ray films is a measure of the difference between the optical density of
adjacent areas in the film (Davidson, 2006). Radiographic contrast can be measured using
densitometers calibrated for the same luminous condition as that of an x-ray film lightbox.
The formation of film optical density is a chemical process. A film’s optical density is a function
of film radiation exposure and is often presented in graph form, which has the shape of a sigmoid
curve (s-curve), known as the characteristic curve. The characteristic curve is divided into a tow,
a latitude (in the middle), and a shoulder. In the latitude, the curve’s slope is approximately straight,
and a film’s radiation exposures within this latitude produces useful film optical contrast. Hence,
a limited range of film radiation exposure is able to produce a desirable optical density.
Radiographic contrast, by its definition, is unchangeable. However, the visual perception of the
radiographic contrast can be optimised to a limited extent by changing the visible luminance level,
The visual perception of x-ray films and x-ray digital images can be optimised by changing the
viewing conditions. The optimum perception of x-ray films or images can be explained by the
Rose (1948) model, which uses the statistics of photon Poisson noise in a background region to
determine an absolute scale for the performance of the human visual system. This model explored
the quantum nature of light to study the limits of human visual system performance. Sturm and
MORGAN (1949) successfully used this model to determine the fundamental role that photon
statistics would play in limiting the clarity of intensification images for visualisation.
15
The Rose (1948) model is valid for low-contrast signals, as pointed out by Burgess (1999).
However, it is only an adequate approximation for determining the optimum visual perception of
in different ways, such as Weber contrast, Michelson contrast, Ratio contrast, or Root mean square
Weber contrast is calculated as the ratio of the luminance of the feature to the luminance of the
feature’s background (Equation 1.1) and is also referred to as ‘Weber fraction’. Weber contrast is
suitable for small features that are present on a large uniform background. It can be used to
determine radiographic contrast of x-ray films (Webber et al., 1977, Koedooder and Venema,
1986).
|𝐼−𝐼𝑏 |
𝐶= (1.1)
𝐼𝑏
where C is the Weber contrast, I, is the luminance of the features, and, Ib is the
luminance of the background.
Michelson contrast is known as the visibility calculated as Equation 1.2. It is commonly used for
patterns where both bright and dark features are equivalent and take up similar fractions of the
area. Michelson contrast is used to determine x-ray image contrast (International Commission on
Radiation Units and Measurements, 1989, Strid, 1976, Sisniega et al., 2013) and is defined as:
16
𝐼𝑚𝑎𝑥 −𝐼𝑚𝑖𝑛
𝐶= (1.2)
𝐼𝑚𝑎𝑥 +𝐼𝑚𝑖𝑛
where C is the Michelson contrast, Imax is the highest luminance of an area, and Imin is
the lowest luminance of this area.
Ratio contrast (Equation 1.3) is the ratio of the transmitted luminance I1 of an area to the
transmitted luminance I2 of this area’s surrounding (Bonenkamp and Hondius Boldingh, 1959a,
Morgan, 1946b):
𝐼1
𝐶= (1.3)
𝐼2
where C is the ratio contrast, I1 is the transmitted luminance of an area, and I2 is the
transmitted luminance of this area’s surrounding.
Root mean square (RMS) contrast (Equation 1.4) is often used in image signal analysis. RMS
contrast is determined by the standard deviation of the pixel intensities (Peli, 1990):
1
𝐶= √ ∑𝑁−1 𝑀−1 ̅2
𝑖=0 ∑𝑗=0 (𝐼𝑖𝑗 − 𝐼 ) (1.4)
𝑀×𝑁
where intensities Iij are the ‘i-th’ and ‘j-th’ element of the two-dimensional image of
size M by N. Ī is the average intensity of all pixel values in the image. The image (I) is
assumed to have its pixel intensities normalised in the range [0, 1].
17
1.2.5.4. Effects of scatter radiation on image contrast
Scatter radiation, when it reaches the image receptor, together with primary radiation, creates a
latent image. During the process of converting the latent image to a digital image, scatter radiation
cannot be separated from primary radiation and therefore cannot be removed. The effects of scatter
radiation on image contrast can be quantified using Webber contrast. When imaging system noise,
which may also affect image contrast, is neglected, the image contrast without scatter radiation is
equal to the subject contrast, which can be calculated by Equation 1.5. When scatter radiation is
|𝐷𝑝𝑟𝑖𝑚𝑎𝑟𝑦 −𝐷𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 |
𝐶= (1.5)
𝐷𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑
|𝐷𝑝𝑟𝑖𝑚𝑎𝑟𝑦+𝑠𝑐𝑎𝑡𝑡𝑒𝑟 −𝐷𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑+𝑠𝑐𝑎𝑡𝑡𝑒𝑟 |
𝐶= (1.6)
𝐷𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑+𝑠𝑐𝑎𝑡𝑡𝑒𝑟
The reduction of image contrast attributed to scatter radiation is a function of scatter radiation to
primary radiation ratios (SPR). In radiography, SPR varies from less than 1 to more than 25 for
imaging anatomy thicknesses of 5–50 cm (Fetterly and Schueler, 2007), assuming the image
contrast without scatter radiation is 0.2. The reduction of this image contrast is illustrated in Figure
1.4. As SPR increases from 0 to 9, this image contrast reduces from 0.2 to approximately 0.02.
18
0.2
(Webber conrast)
Image contrast
0
0 2 4 6 8
SPR
Figure 1.4. Reduction of the image contrast attributed to scatter radiation. The image contrast
reduces as the SPR increases.
Tromans et al. (2010) found the reduction of image contrast attributed to scatter radiation is
irreversible in the image. They attempted to remove the effect of scatter radiation on digital images
and found that the ‘noise’ ascribed to the scatter radiation cannot be separated from the primary
The detrimental effects of scatter radiation on image contrast is permanent and cannot be undone.
Substantially reducing the scatter radiation reaching the image receptor is essential to create
In medical x-ray imaging, x-ray radiation is generated from a finite focal spot. This radiation is
called primary radiation. Scatter radiation is radiation that, during passage through the imaging
The air-gap technique can also reduce scatter radiation reaching the image receptor. In the air-gap
technique, a large space is kept between the imaging object and the image receptor, by moving the
image receptor away from the imaging object. The reduction of scatter radiation increases as this
19
space increases (Sorenson and Floch, 1985). When the space increases, radiation exposures are
increased to compensate for the reduction of primary radiation intensities. This reduction is
ascribed to the distance increased between the focal spot and the image receptor (Neitzel, 1992,
The slot technique reduces scatter radiation reaching the image receptor by keeping the exposing
region within a narrow slot region. This technique is like a single slice CT in which the primary
radiation beam is confined to form a narrow (for example, 1 cm) fan beam. Then the fan beam is
guided over the patient during radiation exposure (Jaffe and Webster, 1975, Barnes et al., 1976,
Barnes et al., 1993). The radiation exposure time for using the single slot technique is extremely
long because of the limited coverage of a single fan beam. X-ray tube heat loading is approximately
increased by the ratio of the imaging size to the slot width but can be reduced by using multiple
slots. The high tube heat loading, complex mechanical control system of multiple slots, long
radiation exposure time, and difficulty in controlling radiation exposure time makes the
Current attempts to reduce scatter radiation focus on software solutions (Mentrup et al., 2014,
Takahiro Kawamura et al., 2015, Renger et al., 2016). These attempts study scatter radiation
distribution and subtract an estimate of scatter radiation to obtain a new image. Some applications
of these attempts have been found, such as in bedside x-ray examinations. They improve image
quality, but not as well as the anti-scatter grid technique. Furthermore, these attempts have a
fundamental challenge in that the scatter radiation distribution of an anatomical region may vary
between individuals. An inaccurate estimation of scatter radiation might introduce artefacts, which
20
In 1913, Bucky (1913) introduced a grid in between an image receptor and an imaging object to
take an x-ray image of that object. He found that the improved clarity of this image was attributed
to the reduction of scatter radiation by the grid. The technique is known as an anti-scatter grid
technique and is the most common technique used in medical x-ray imaging systems to reduce
Anti-scatter grids consist of a periodic array of radiopaque foil strips separated by radiolucent
material that fills the strip interspaces. The mechanism of scatter radiation reduction by grids was
first described by Bucky (1913). The description of this mechanism can be found in several
textbooks, such as Bushberg et al. (2011), Carlton et al. (2012), and Bushong (2013).
An illustration of an anti-scatter grid’s mechanism is given in Figure 1.5. After primary radiation
is generated at the focal spot, the primary radiation is confined to a shape, such as a rectangular
shape, and is directed to the irradiated object. Some primary radiation travels through the object
and reaches the grid. Some primary radiation is scattered by the object’s matter to produce scatter
radiation which may reach the grid. Some primary radiation is absorbed in the object’s matter.
Most scatter radiation reaching the grid is absorbed in the radiopaque strips. Most primary
radiation reaching the grid travels through the radiolucent interspaces that are aligned to the
primary radiation’s direction. Thus, most scatter radiation is stopped from reaching the image
receptor, and most primary radiation travels through the grid and reaches the image receptor.
21
Figure 1.5. Reduction of scatter radiation by an anti-scatter grid. Some primary radiation and
some scatter radiation travel through the grid and reach the image receptor. Some primary
radiation and some scatter radiation are absorbed in the grid’s strip materials.
There are different types of grids, such as focused grids, parallel grids, cross-hatch grids, and
tapered grids. Focused grids are linear grids with their strips aligned to a common line focus and
are the most common type of grids implemented in medical x-ray imaging systems.
Grid strips can be made from materials of high atomic number, such as tungsten or lead. High
atomic number materials absorb a great fraction of radiation travelling through them. The
interspace materials can be made from radiolucent materials, such as aluminium, polymethyl
methacrylate (PMMA), wood, or carbon fibre. These materials let a great fraction of radiation
22
1.2.7.2. Reduction of primary and scatter radiation by grids
The reduction of primary radiation and scatter radiation in grid materials can be quantified as the
transmissions of primary and scatter radiation. The transmission of primary radiation (Tp) is the
ratio of primary radiation reaching the image receptor with a grid in place to that without a grid in
place. The transmission of scatter radiation (Ts) is the ratio of scatter radiation reaching the image
In the scholarly literature, the Tp of carbon-interspaced grids decreases from 0.764 to 0.640 as the
grid ratio increases from 8:1 to 21:1 (Fetterly and Schueler, 2007, Fetterly and Schueler, 2009). A
similar trend in Tp was shown in Kim et al. (2007), in which Tp decreases from 0.76 to 0.71 as the
grid ratio increases from 5:1 to 10:1. Generally, Tp varies from 0.64 to 0.77 with a mean of
approximately 0.72 (SD = 0.034) (Fetterly and Schueler, 2007, Kim et al., 2007, Fetterly and
Schueler, 2009) for general grids or from 0.74 to 0.80 with a mean approximately 0.78 (SD =
0.038) for mammographic grids (Carton et al., 2009, Salvagnini et al., 2012).
Ts decreases from 0.235 to 0.058 as the grid ratio increases from 8:1 to 21:1. A similar trend in Ts
was also shown in Kim et al. (2007), in which Ts decreases from 0.40 to 0.30 as the grid ratio
increases from 5:1 to 10:1. In general, Ts varies from 0.058 to 0.456 with a mean of approximately
0.212 (SD = 0.142) (Fetterly and Schueler, 2007, Kim et al., 2007, Fetterly and Schueler, 2009)
for general grids or from 0.125 to 0.196 with a mean of approximately 0.156 (SD = 0.046) for
Grids reduce a large fraction of scatter radiation and a small fraction of primary radiation reaching
the image receptor. For example, a typical grid ratio 10:1 grid has Tp equal to 0.75 and Ts equal
to 0.17 (Fetterly and Schueler, 2007). When this grid is used in an x-ray examination of 20-cm
23
thick abdomen, which has an SPR of approximately 6:1 (Fetterly and Schueler, 2007), (Fetterly
and Schueler, 2007), the relative amount of scatter radiation to the amount of primary radiation
reaching the image receptor would be approximately 1.34:1. As illustrated in Section 1.2.5, this
relative amount scatter radiation is a major degradation factor reducing the image contrast.
As photons carry specific amounts of energy and exhibit the properties of waves and particles
(Carlton et al., 2012), the photons’ statistics can be determined using Poisson distribution which
governs the statistics of random events that are independent of each other and time (International
Atomic Energy Agency, 2014). Poisson statistical variance of photons can be measured in the
coefficient of variation (COV) for COV = 1/sqrt(A), where ‘A’ is the number of photons and sqrt(A)
is the square root of ‘A’ (International Atomic Energy Agency, 2014, Bushberg et al., 2011). The
Grids reduce the number of primary photons and hence increase these photons’ COV. The effect
of this increase on the x-ray image can be perceived as Poisson mottle (Grangeat, 2009). An x-ray
image’s Poisson mottle is reduced by increasing radiation exposure, which determines the number
An increase in radiation exposure, which is ascribed to the compensation for the reduction on the
primary radiation in the grid materials, increases patient radiation exposures. This increase is
approximately 39% for general grids (mean Tp = 0.72) and 28% for mammographic grids (mean
24
1.3. Research opportunity from the literature
The research opportunity for this project emanates from the scholarly literature with regard to two
areas:
1) Using grids fails to minimise radiation exposure delivered to patients because of the
compensation for the reduction of primary radiation in grid materials.
2) Using grids lacks adequate optimisation of image diagnostic quality because of the
remaining scatter radiation in the image.
1.4. Motivations
Grids have been used in x-ray imaging systems since the 1920s. Contemporary grids have average
Tp and Ts equal to 0.72 and 0.211 for general grids and 0.78 and 0.156 for mammographic grids,
respectively. Without using grids, especially x-ray images of anatomies thicker than 20 cm, images
will not be created with adequate contrast for making sensible diagnoses. When grids are used, the
stochastic effects are increased by 39% in general radiography and 28% in mammography.
This project was motivated by the fact that the use of grids will not change in the coming decades.
If primary radiation in grids is not reduced, increases in the stochastic effect attributed to using
1.5. Purpose
The primary purpose of this project is to investigate anti-scatter grids and their Tp and Ts. In
addition, this project will also associate these Tp and Ts with the stochastic effects of ionising
radiation as a stochastic effect reduction management strategy the implementation of the ALARA
principle.
25
1.6. Significance
This project will make a significant and original contribution to the scholarly literature and
provides a potential solution in reducing stochastic effects. This solution will also be useful for
medical x-ray imaging systems, particularly in the areas of dose-image quality optimisation. First,
a significant contribution relates to the reduction of stochastic effects as a means of addressing the
problem of the radiation transmissions in grids, as a radiation dose management strategy, and as a
practical and important option in the implementation of the ALARA principle. Second, as an
Finally, this project is worthy of research ascribed to the recent increasing concerns regarding the
stochastic effects of low-dose ionisation radiation exposure from medical imaging (Schultz-Hector
and Trott, 2007, Little et al., 2008b, Little, 2010, Baker et al., 2011, Picano et al., 2012). When
radiation doses fall, stochastic effects fall proportionally. However, as there is no threshold dose,
this means that any radiation dose, no matter how small, has the potential to cause harmful effects,
such as cancer or leukaemia. The chance of these harmful effects occurring increases
The use of grids, which are of great benefit in patient disease management, increases the stochastic
effects of ionising radiation and is a serious and important problem, for which solutions have been
sought since 1913 (Wilsey, 1923, Morgan, 1949, Bonenkamp and Hondius Boldingh, 1959b, Strid,
1976, Kalender, 1982, Chan et al., 1985, Fetterly and Schueler, 2007, Fetterly and Schueler, 2009,
Cunha et al., 2010). A solution to this problem will be a significant step in the quest to reduce
ionising radiation delivered to patients without compromising image diagnostic quality. Thus, this
26
solution will minimise the probability of stochastic effects on patients undergoing x-ray
examinations.
This research has three distinct phases. In the first phase, this project will quantitatively evaluate
Tp and Ts. As the manufacture of many prototype grids is not only expensive but requires a lot of
time, experimental evaluation of many prototype grids is not feasible in this project. An alternative
grid evaluation method, the Monte Carlo simulation, will be developed and validated. In the second
phase, this project will explore the relationship between Tp and Ts. A new grid design method will
be established, which can assist in designing new grids with Tp increased and Ts reduced
concurrently. In the third and final phase, many new grids will be designed and evaluated in the
(5) to evaluate new grids using the Monte Carlo simulation code system.
27
1.9. Research contributions
The main goal of this research is to investigate grids, to address radiation transmission in grids,
and to associate these transmissions with the stochastic effects of radiation exposure as a stochastic
effect reduction management strategy in the ALARA principle. This goal was achieved through a
series of actions performed in three distinct phases and through original contributions to the
(1) A new grid design method: This new grid design method provides criteria to design grids
with negligible reduction of primary radiation and without compromising the reduction of
scatter radiation. This method solved the problems of the reduction of primary radiation in
grid materials. In other words, this method provided a solution for designing new grids with
the expectation that this will improve the transmission of primary radiation, for example, Tp >
0.99, without compromising the reduction of scatter radiation.
(2) A novel criterion to determine the strip optimal thickness: This novel criterion employs
the first differential of the quantum signal-to-noise ratio improvement factor (KSNR) to
determine a condition for the determination of strip optimal thicknesses. This criterion
assumed that if the maximum KSNR exists, it will exist at the zero value of its first differential.
(3) A new method for determining radiation transmissions in grid materials: This new
method for determining of radiation transmissions in grid materials was used in the Monte
Carlo simulation to determine radiation transmissions in grid materials. This method
overcame the limitations of those set out in the literature. The development and validation of
this method has been published (Zhou et al. (2016).
(4) A new Monte Carlo simulation code system: The new Monte Carlo simulation code system
can be used to simulate x-ray imaging systems. This system was developed in Matlab Ver.
R2015b (The MathWorks Inc., Natick, Massachusetts) and validated for the evaluation of
grids. This system has the advantage of simulating millions of photons’ transportation
simultaneously by using the array operation in the MatLab software environment. Validation
28
of this Monte Carlo code system was presented at the Engineering and Physical Science in
Medicine 2016 conference and is in preparation for publication.
(5) Many new grid designs: Many new grid designs were created and evaluated in this project
through Monte Carlo simulation. The application of new grids from these new designs will
reduce negligible amounts of primary radiation without sacrificing the reduction of scatter
radiation. These new grids can supersede current grids and be used as a stochastic effect
reduction management strategy.
This thesis is organised into eight chapters. Chapter 1 introduced the nature and scope of this study
by identifying, in the scholarly literature, problems with contemporary grids, the need for a
Chapter 2 will review methodological approaches that can be used in this project to conduct
investigate a novel criterion using the first differential of KSNR to determine the strip optimal
Chapter 3 will elucidate information for the development of a new Monte Carlo simulation code
system for simulating x-ray imaging. These principles will be used to guide the development of a
new Monte Carlo simulation code system, which will be used in this project to evaluate many new
grid designs.
Chapter 4 will clarify the simulation of photon transportation in grid materials. It will first review
methods for determining radiation transmissions in grid materials and then explain the limitations
29
of these methods. Following that, this chapter will continue to establish and validate a new method
Chapter 5 will present the validation of the new Monte Carlo code system developed using the
Chapter 6 will focus on a new grid design method to overcome the reduction of primary radiation
in grid strips. This chapter will first present an overview of current grid design, will then discuss
grid design factors and issues, and finally will examine the criteria for designing new grids.
Chapter 7 will apply the achievements made in Chapters 2 to 6 to determine a solution for the
reduction of primary radiation in grid strips. This chapter will present many new grid designs
Chapter 8 will present the interpretation of the results made in this project and draw conclusions
from these results. Furthermore, recommendations for future research and summary will be
30
Chapter 2: Grid evaluation methodology
This chapter marks the start of the first phase and highlights a methodology for quantitative
evaluation of the transmissions of primary (Tp) and scatter (Ts) radiation. It has five main sections:
the first reviews criteria for grid evaluation; the second proposes a novel criterion for determining
strip optimal thicknesses; the third focuses on radiation measurements in grid evaluation; the fourth
presents the setups for radiation measurements in the experiment; and the last discusses the
Grid evaluation is used to quantify the reduction of scatter radiation, the reduction of primary
radiation, and their effects on image contrast. Scatter radiation reduces image contrast and leads
to difficulty in making accurate diagnoses whereas primary radiation carries information about
anatomical structure to form a useful image. The amount of reduction of scatter and primary
radiation are used to derive grid performance criteria. New grid criteria for digital imaging
Approximately four decades after Bucky (1913) invented anti-scatter grids to improve x-ray image
clarity, Bonenkamp and Hondius Boldingh (1959a) carried out systematic investigations of grid
performance and presented grid performance criteria; these are widely used today to determine
grid performance.
31
The basic grid performance criteria are Tp, Ts, and the transmission of total radiation (Tt). All
three are fundamental quantities that indicate grid performance and are essential quantities for
The derived grid performance criteria are the image contrast improvement factor (Morgan, 1946a,
Bonenkamp and Hondius Boldingh, 1959a), the Bucky factor, the quantum signal-to-noise ratio
improvement factor (Dick and Motz, 1978, Neitzel, 1992), and the image quality improvement
factor (International Electrotechnical Commission, 2013). The basic criteria and these derived
Tp, Ts, and Tt are characteristics of grids. Tp is the ratio of the amount of primary radiation with
a grid to without a grid and is calculated using Equation 2.1. Ts is the ratio of the amount of scatter
radiation with a grid to without a grid and is calculated using Equation 2.2. Tt is the ratio of the
amount of total radiation with a grid to without a grid and is calculated using Equation 2.3. Tp, Ts,
and Tt are used to derivate other grid performance factors, of which common ones are further
discussed below.
𝑃𝑟𝑖𝑚𝑎𝑟𝑦_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ 𝑔𝑟𝑖𝑑
𝑇𝑝 = (2.1)
𝑃𝑟𝑖𝑚𝑎𝑟𝑦_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑔𝑟𝑖𝑑
𝑆𝑐𝑎𝑡𝑡𝑒𝑟_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ 𝑔𝑟𝑖𝑑
𝑇𝑠 = (2.2)
𝑆𝑐𝑎𝑡𝑡𝑒𝑟_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑔𝑟𝑖𝑑
𝑇𝑜𝑡𝑎𝑙_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ 𝑔𝑟𝑖𝑑
𝑇𝑡 = (2.3)
𝑇𝑜𝑡𝑎𝑙_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑔𝑟𝑖𝑑
32
2.1.3. Image contrast improvement factor: K
The image contrast improvement factor (K) is also known as the image contrast improvement ratio.
primary radiation and a small fraction of scatter radiation. In film-screen imaging systems, the
transmitted primary radiation carries anatomical information while the transmitted scatter radiation
helps ensure desirable film optical density. K is an acceptable measure of the compromise between
the film image diagnostic quality and patient radiation dose (Bonenkamp and Hondius Boldingh,
𝑇𝑝
𝐾= (2.4)
𝑇𝑡
In film-screen systems, an increase in radiation exposure to compensate for this fraction is needed
to ensure that the film’s radiation exposure is within the film’s latitude (discussed in Section
1.2.5.2). In digital imaging systems, an increase in radiation exposure to compensate for the
reduction of primary radiation is needed to avoid increasing the image Poisson mottle (discussed
in Section 1.2.7.2).
The relative increase in radiation exposure is described by the grid Bucky factor (B), which
accounts for the compensation for the reduction on both primary radiation and scatter radiation.
33
This factor is useful for screen-film systems but not for digital imaging systems (Fetterly and
1
𝐵= (2.5)
𝑇𝑡
image detectors (Dick et al., 1978). Digital images’ contrast and brightness can be altered through
digital image processing algorithms (referred to in Section 1.2.5). A digital image’s diagnostic
quality mainly depends on the image’s SNR, and the SNR depends primarily on the ratio of
primary radiation to scatter radiation. KSNR is a measure of a digital image’s SNR improvement
(Dick and Motz, 1978, Neitzel, 1992, Fetterly and Schueler, 2007) and is calculated using Equation
2.6.
34
𝑇𝑝2
𝑄= = 𝑇𝑝 × 𝐾 (2.7)
𝑇𝑡
where Tp and Tt are the transmissions of primary and total radiation, respectively; K
is the image contrast improvement ratio; and Q is the image quality improvement factor.
approach one and KSNR approaches one. As the strip thickness approaches infinity, Tp, Ts, and Tt
approach zero and KSNR is likely approaching zero because the elements in the right side of
Equation 2.8 are approaching zero. However, the KSNR of grids as set out in the scholarly literature
is generally greater than one. The KSNR of a carbon-fibre-interspaced grid with a grid ratio of 8:1
is about 1.37 (Fetterly and Schueler, 2007). These KSNR values suggest that KSNR has a maximum
𝑃𝑔+𝑆𝑔 𝑇𝑝×𝑃+𝑇𝑠×𝑆
𝐾𝑆𝑁𝑅 = 𝐾 √𝑇𝑡 = 𝐾 √ = 𝐾√ (2.8)
𝑃+𝑆 𝑃+𝑆
Strip optimal thickness is defined as the strip thickness with which, for a given grid ratio, strip
height, and cover material and thickness, the grid has the maximum KSNR. Directly finding the
35
strip optimal thickness can be problematic, but it can be indirectly determined by finding the
maximum KSNR.
KSNR can be expressed as a function of grid ratio, strip height, strip thickness, cover thickness, and
the linear attenuation coefficients of the strip, cover, and interspace materials (Equation 2.9). For
a given grid ratio, strip height, strip and interspace materials, cover material and thickness, the
maximum KSNR can be found at the condition in which the first differential of Equation 2.9, with
respect to the strip thickness, is equal to zero (Equation 2.10). After the maximum KSNR is found,
where, r is grid ratio; h is strip height; d is strip thickness; cth is cover thickness; μs, μi,
and μc are the strip, interspace, and cover materials’ linear attenuation coefficients,
𝛿𝑦
respectively; 𝛿𝑥 is the first differential of function y with respect to variable x; and KSNR is
the quantum signal-to-noise ratio improvement factor.
The amount of radiation measured is proportional to the amount of signal that is produced when
the radiation interacts with the detector materials. This signal depends on the amount of energy
deposited in the detector materials. The energy deposited in a material depends on three factors:
the density of the material, the mass energy absorption coefficients, and the thickness of the
36
As x-ray beams used in x-ray imaging are usually poly-energetic and have a spectrum of energy,
the energy deposition in the detector materials can be specified at each photon energy or expressed
as an ‘effective’ value over the beam spectrum reaching the detector. The x-ray beam spectrum
depends on the peak kilovoltage (kVp) applied to the generator and the characteristics of the
generator (such as anode material, anode angle, and filtrations). The x-ray beam spectrum is also
influenced by, if presented, the imaging anatomy which attenuates, or ‘hardens’, the beam and
The amount of radiation energy deposited in a detector depends on the material thickness, mass
density, and mass energy absorption coefficients of the detector. The mass energy absorption
coefficient generally is higher at lower photon energy and decreases as photon energy increases.
However, if the material has an atomic absorption edge, which is always true for high atomic
materials (e.g. Gadolinium has absorption edges at 50.24 kilo-electron-volt (keV) and 8.37 keV),
the energy deposition increases dramatically above the absorption edge, which causes a local
The amount of signal produced by photon energy deposition in a detector depends on the
interactions between the photons and the detector materials. Though photons may interact with
matter through Rayleigh scattering, Compton scattering, or photoelectric absorption, the main
interaction with the material of a high atomic number detector is photoelectric absorption. In
photoelectric absorption, the interaction of a photon with the detector materials causes an electron
to be emitted from its atom. In the electron’s subsequent loss of kinetic energy, excitation and
ionisation occur and produce secondary signals, such as optical photons or electronic charges.
37
The amount of radiation can be measured in different quantities, such as total energy, absorbed
dose, photon fluence, and radiation energy fluence. For x-ray images, ideally all energy in the
beam should be converted to an image signal, but many radiation quantities do not measure the
total energy of the beam. The absorbed dose measures energy deposition characterised by the mass
energy absorption coefficient but the absorbed dose is not a measure of the total energy of the
beam. Photon fluence is a measure of the total number of photons per unit area but does not account
for the photon energy. Radiation energy fluence is a measure of the total energy per unit area and
In grid evaluation, radiation measurements are made with a variety of detectors, for example x-ray
film (Wilsey, 1921), Ge detector (Ter-Pogossian et al., 1974), gas ionisation chamber (Strid, 1976),
NaI crystal (Barnes et al., 1976, Dick et al., 1978), fluorescent screen with photo-detector
(Bonenkamp and Hondius Boldingh, 1959a, Niklason et al., 1981), and photostimulable storage
phosphor plate (Floyd et al., 1991, Flynn and Samei, 1999, Fetterly and Hangiandreou, 2001,
Fetterly and Schueler, 2007, Fetterly and Schueler, 2009). The International Electrotechnical
As this screen is not commercially available anymore, it has been replaced with gadolinium
detector, lights that are emitted from the fluorescent screen are measured. The results of radiation
transmission then depend on the detection efficiency of the photo-detector and the light-conversion
efficiency of the fluorescent screen, which is a function of the screen area density and radiation
beam quality.
38
In grid evaluation, differences between radiation detectors could not be excluded from affecting
grid evaluation results. Floyd et al. (1991) found that scatter fraction measured with
photostimulable storage phosphor plates was in acceptable agreement with those measured with
films and most other previously reported measurements. Floyd et al. (1991) did not account for
differences in experimental setups, such as phantom sizes, primary beam’s field of view (FOV),
and kVp. These differences have been found to affect the distribution of SPR (Boone and Seibert,
approximately the same as the SPR in Fetterly and Schueler (2007) for the 30 cm × 30 cm × 20
cm (W × L × H) phantom. Theoretically, these SPR should be different from each other. Difficulty
in experimental setups has been found to influence grid evaluation results (Fetterly and Schueler,
In the Monte Carlo simulation, correction can be made for radiation measurements from different
detectors which depend on x-ray beam quality and photons’ entry angles. Shuping and Judy (1977)
found that for a constant radiation exposure (measured in air-absorbed dose), the energy deposited
in a fluorescent screen increased as the tube kVp increased. This result shows that fluorescent
screens yield different amounts of light for the same air-absorbed dose of different x-ray beam
quality. Photons’ entry angles to the detector entry surface affect the photons’ travelling distances
in the materials of the detector. These distances affect the probabilities of interactions between
photons and the materials in the detector. Hence, these angles affect the photons’ energy deposition
39
2.4. Experimental setup for grid evaluation
Grid evaluation requires the measurement of scatter, primary, and total radiation with and without
The amount of scatter radiation can be quantified with a modulation transfer function (MTF)
method (Salvagnini et al., 2012). MTF measures how accurately an imaging system reproduces
the details of an object to its image (Boreman, 2001). In x-ray imaging, MTF is a measure of the
transfer of the subject contrast to the image contrast. Boone and Seibert (1988) demonstrated the
effect of scatter radiation on image contrast by comparing MTF with and without scatter radiation.
Saunders Jr and Samei (2006) further examined the effect of scatter radiation on MTF using Monte
Carlo simulation and found a similar result as Boone and Seibert (1988). Their result showed that
scatter radiation reduces low frequency in MTF. Salvagnini et al. (2012) extended this result to
quantify scatter radiation using the reduction of low frequency in MTF. The MTF method requires
In the beam stop method, measurements of the amount of primary, scatter and total radiation were
performed under three conditions: a narrow beam condition, a broad beam condition with a
primary beam blocker, and a broad beam condition (Bonenkamp and Hondius Boldingh, 1959a,
The amount of primary radiation was measured in the narrow beam condition with the radiation
beam tightly confined to the radiation detection area with diameter of approximately 6 mm (Figure
2.1). The phantom (water phantom for general grids, and Polymethyl methacrylate (PMMA)
phantom for mammographic grids) was placed as close as possible to the x-ray source/focal spot.
40
Lead diaphragms were placed above and below the phantom. The radiation beam entering the
phantom and reaching the detector only travelled through the area confined by these diaphragms.
Another set of lead diaphragms was placed in between the detector and the phantom to further
The radiation beam was centred to the grid which was placed at its focal distance from the x-ray
source. The detector was placed underneath the grid without a space. Lead diaphragms were placed
underneath the detector to prevent back scatter radiation. The primary radiation was then measured
Figure 2.1. Arrangement of x-ray source, phantom, grid, and detector in the narrow beam
condition. (a) Front setup view of general grid evaluation. (b) Side setup view of
mammographic grid evaluation.
41
2.4.2. Broad beam condition with a primary beam blocker
The amount of scatter radiation was measured in the broad beam condition with a primary beam
blocker. In this condition, the centre of the primary beam was blocked by a lead blocker (Figure
2.2). The grid was placed at its focal distance from the x-ray source/focal spot. The detector was
placed underneath the grid without a space. Lead diaphragms were placed underneath the detector
The phantom (water phantom for general grids, and Polymethyl methacrylate (PMMA) phantom
for mammographic grids) was placed above the grid and as close as possible to it. The projection
of the lead blocker was to just cover the entire radiation detection area with a diameter of
approximately 6 mm. The thickness of the lead blocker was adequate to substantially stop primary
radiation reaching the detection area. The primary beam was confined to an area of 30 cm × 30 cm
at the phantom exit-surface for general grid evaluation or 15 cm × 15 cm for mammographic grid
evaluation. The scatter radiation was then measured with and without the grid.
42
Figure 2.2. Arrangement of x-ray source, phantom, grid, and detector in the broad beam
condition with a blocker. (a) Front setup view of general grid evaluation. (b) Side setup view of
mammographic grid evaluation.
The setup for the broad beam condition (water phantom for general grids, and Polymethyl
methacrylate (PMMA) phantom for mammographic grids) was the same as for the broad beam
condition with a lead blocker, except that the lead blocker was removed (Figure 2.3). The amount
43
Figure 2.3. Arrangement of x-ray source, phantom, grid, and detector in the broad beam
condition. (a) Front setup view of general grid evaluation. (b) Side setup view of
mammographic grid evaluation.
The simulation setups for grid evaluation can be made the same as the experimental setups. These
setups do not account for the forward scatter radiation that arises in the narrow beam condition,
nor do they account for the scatter radiation that otherwise would arise in the phantom where
primary radiation is blocked in the broad beam condition with a lead blocker. These setups,
however, would require two times more simulation time because the measurements require three
different setups.
In this project, the simulation setup was made the same as the broad beam condition. With this
setup, primary radiation and scatter radiation were labelled by primary and scatter markers. The
energies of photons distinguished by markers were tallied for the calculations of Tp and Ts. This
44
setup accounted for the forward scatter radiation arising in the primary beam’s FOV in the narrow
beam condition and setup accounted for the scatter radiation which otherwise would not arise in
Summary
This chapter examined the criteria for grid evaluation with an emphasis on the basic criteria: Tp,
Ts, and Tt. These basic criteria were used to calculate the derivative grid evaluation criteria. The
derivative grid evaluation criterion, KSNR, was discussed for the determination of grid strip optimal
thickness. A novel criterion, the first derivative of KSNR, was proposed to determine the optimal
Furthermore, this chapter reviewed the difference between the measurements of the amount of
radiation in grid evaluation and highlighted the beam-block method for the measurements of
primary, scatter, and total radiation. This beam-block method determined primary, scatter, and
total radiation using three measurement conditions: narrow beam condition for primary radiation,
broad beam condition with a primary beam blocker for scatter radiation, and broad beam condition
for total radiation. In addition, this chapter also discussed the differences in radiation measurement
These grid evaluation criteria and the beam-block method for radiation measurement were used in
the experiment to determine grid performance or in the Monte Carlo simulation to determine grid
design performance. The next chapter will discuss the development of our Monte Carlo simulation
45
Chapter 3: Information for the development of the
Monte Carlo code system
This chapter presents the second part of the first phase in this project. It discusses the simulation
of x-ray imaging and presents an overview of Monte Carlo simulation. It then delineates a
procedure for simulating the production of x-ray images. After that, the chapter details photon
transportation and interactions between photons and matter. Finally, the radiation dosimetry and
imaging phantom construction used in the Monte Carlo simulation are discussed. A dedicated
Monte Carlo simulation code system will be developed using the information discussed in this
chapter.
3.1. Introduction
Monte Carlo simulation was first used for exploring the behaviour of neutron chain reactions in
fission devices in the 1940s. At that time, physicists were unable to understand the behaviour of
Stanislaw Ulam and John von Neumann proposed a statistical and random sampling technique to
solve problems like the behaviour of neutron chain reactions. Simulation codes were developed
using this statistical and random sampling technique, and these codes were successfully executed
on an electronic computing machine at the Los Alamos Scientific Laboratory (Eckhardt, 1987).
Monte Carlo simulation has since been used to resolve a wide range of complex physical and
mathematic problems. In Monte Carlo simulation, a problem can be fit into a model in which an
event happens in a random manner that depends on a sequential sample of random numbers. For
47
different sequences of random numbers, the simulation does not yield identical results; however,
they agree with each other within some errors due to statistic variation (Landau and Binder, 2009).
Monte Carlo simulation can virtually replicate x-ray imaging technologies to simulate the
production of x-ray images. In addition, radiation energy deposited in the imaging phantom can
represent an anatomical structure, radiation energy deposited in this phantom can be used to
Monte Carlo simulation is a preferable method used to study x-ray imaging systems. It has been
used to investigate scatter radiation (Kalender, 1981, Jing et al., 1998, and Boone and Cooper III,
2000), evaluate anti-scatter grid performance (Kyriakou and Kalender, 2007, Jang et al., 2008,
Lazos and Williamson, 2010, Tanaka et al., 2013), assess patient radiation dose (Kim et al., 2010,
Cunha et al., 2010), analyse image diagnostic quality (Tomal et al., 2013, Chen et al., 2015), and
validate imaging technologies’ implementation (Yu et al., 2009, Yan et al., 2010, Bornefalk et al.,
The procedure for simulating the production of an x-ray image started with the generation of x-ray
photons at the x-ray focal spot and finished with the deposition of photon energy in the image
receptor. An x-ray imaging system with three basic imaging components (the x-ray generator, the
48
Figure 3.1. Three basic components of an x-ray imaging system.
During the acquisition of an image, primary photons were produced at the focal spot. They were
constrained to travel towards the imaging object. When travelling inside the imaging object, some
photons might be absorbed in the imaging object’s matter, some might travel through the imaging
object without any interaction and reached the image receptor, and some might interact with the
imaging object’s matter and change their direction to become scatter photons. Some scatter
photons, in their subsequent interaction with the matter, might be absorbed in the matter, whereas
some might successfully travel through the imaging object and reach the image receptor.
The intensities of all photons that had reached the image receptor formed the image. Generally,
simulating the production of an image followed the tracks of photons from the x-ray focal spot to
49
Following a photon’s track is just like transporting the photon from the focal spot to the image
receptor, or to where the photon energy is completely absorbed, or to where the photon meets
predefined conditions. The transportation of photons in the simulation of an x-ray imaging system
2) Determine the photon’s properties: energy, position, direction, and status (primary
or scatter)
4) Transport the photon by the distance of the free path length to a new location
(A) Yes.
a. If Rayleigh scattering
iii. Go to step 3
b. If Compton scattering
50
ii. Determine the photon’s new energy and direction
iv. Go to step 3
c. If photoelectric absorption
ii. Go to step 9
(B) No.
i. Go to step 6
6) Transport the photon to the plane in which the detector’s detection area is
51
Figure 3.2. Flow chart of photon transportation in Monte Carlo simulation
52
3.3. Photon transportation
Photon transportation included changes in a photon’s location, direction, energy, and status
(primary or scatter). The change in location was determined using a free path length, which is the
distance that a photon travels without any interaction with matter. A free path length was a random
length determined by the linear attenuation coefficient and a random number. The change in
direction depended on the interaction type and two random angles. Each of these changes is
space. A three-dimensional Cartesian coordinate system was defined by x-, y- and z-axes. A
photon’s location and direction were represented by a vector, which was determined using six
variables: coordinates (x, y, z) and direction cosines (α, β, γ). A constant relationship between
variables α, β, and γ is given in Equation 3.1. The positive z-axis direction was defined from the
The initial coordinates (x, y, z) of a photon were sampled from the x-ray focal spot area, where
photons are generated. The initial direction cosines (α, β, γ) of this photon were sampled in a
probability distribution function, such as an isotropic direction distribution. When this photon
changed its location by a free path length l, the new coordinates (x’, y’, z’) were calculated using
𝛼 2 + 𝛽2 + 𝛾 2 = 1 (3.1)
𝑥 ′ = 𝑥 + 𝑙𝛼 (3.2)
53
𝑦 ′ = 𝑦 + 𝑙𝛽 (3.3)
𝑧 ′ = 𝑧 + 𝑙𝛾 (3.4)
When the direction of a photon was changed, the new direction was determined using the direction
cosines and two random angles: a polar angle θ and an azimuth angle ω (Figure 3.3). The polar
angle was sampled from the interval (0, π]. The azimuth angle was uniformly sampled from the
interval (0, 2π]. All angles were measured with angle units in radians.
If α and β of the direction cosines (α, β, γ) were equal to zero (α = 0 and β = 0), the new direction
cosines (α’, β’, γ’) were calculated using Equations 3.5, 3.6, and 3.7, similar to Salvat (2015p. 36).
54
If α and β of the direction cosines (α, β, γ) were not equal to zero, a polar angle (θ0) and an azimuth
angle (ω0) were calculated using Equations 3.8 and 3.9. The new direction cosines (α’, β’, γ’) were
calculated using Equations 3.10, 3.11, and 3.12. These equations are different from Salvat (2015p.
37) by an angle shift determined by the polar angle θ0. This polar angle θ0 may have a significant
effect on the probability distribution of scatter photons’ directions because the random polar angle
θ is not uniformly sampled. As the azimuth angle ω is uniformly sampled, it does not have any
𝛾 ′ = 𝑐𝑜𝑠(𝜃) (3.7)
𝜃0 = 𝑎𝑐𝑜𝑠(𝛾) (3.8)
𝛼
𝑎𝑐𝑜𝑠 ( ) 𝛽>0
√𝛼 2 +𝛽 2
𝛼
𝜔0 = − 𝑎𝑐𝑜𝑠 ( ) 𝛽<0 (3.9)
√𝛼 2 +𝛽 2
𝛼
𝑎𝑐𝑜𝑠 ( ) 𝛽 = 0, 𝛼 ≠ 0
{ √𝛼 2 +𝛽 2
𝑠𝑖𝑛(𝜃+𝜃0 )
𝛼 ′ = 𝛼 𝑐𝑜𝑠(𝜃 + 𝜃0 ) + (𝛼𝛾 𝑐𝑜𝑠(𝜔 + 𝜔0 ) − 𝛽 𝑠𝑖𝑛(𝜔 + 𝜔0 )) (3.10)
√1−𝛾2
𝑠𝑖𝑛(𝜃+𝜃0 )
𝛽 ′ = 𝛽 𝑐𝑜𝑠(𝜃 + 𝜃0 ) + (𝛽𝛾 𝑐𝑜𝑠(𝜔 + 𝜔0 ) + 𝛼 𝑠𝑖𝑛(𝜔 + 𝜔0 )) (3.11)
√1−𝛾2
55
𝛾 ′ = 𝛾 𝑐𝑜𝑠(𝜃 + 𝜃0 ) − 𝑠𝑖𝑛(𝜃 + 𝜃0 ) 𝑐𝑜𝑠(𝜔 + 𝜔0 ) √1 − 𝛾 2 (3.12)
When a photon changed its location and/or direction, the photon’s new location was calculated
from a free path length. The new direction of this photon was calculated from its direction cosines,
a polar angle and an azimuth angle that was uniformly sampled. The determination of the free path
A free path length (or free path) is the distance that a photon would travel in matter without any
interaction with the matter. The free path length was determined using the linear attenuation
coefficient and a random number. It could also be determined using the interaction cross-section
The interaction cross-section measures the same probability of interactions between x-ray radiation
and matter, as does the mass attenuation coefficient but in different types of units. The interaction
cross-section, the linear attenuation coefficient, and the mass attenuation coefficient all depend on
both the photon’s energy and its mass. The linear attenuation coefficient depends on mass density,
which can vary for a given element or compound. In contrast, the interaction cross-section and the
object layer. This cross-section is a fictitious projection of a particle that can be better understood
by conceiving the particles like tiny spheres. No matter from which direction you approach the
spheres, you will always see the projections of the spheres as cross-sections through the centres of
the spheres. The collision between the photon and the electron can only be possible within the
56
electron’s fictitious cross-section. Each electron has a cross-section multiplied by the total number
The interaction cross-section data (also known as the atomic cross-section data) were calculated
and presented in units of cm2/atom or barns/atom (or b/atom, where 1 barn = 10–24 cm2) (Chantler,
2000). The connection between the interaction cross-section and the mass attenuation coefficient
is the conversion factor, A/NA, where A is the relative atomic mass of the target element and NA is
Avogadro’s number = 6.0221415 x 10–23 atoms mol-1 (Hubbell, 2006). As the connection between
the linear attenuation coefficient and the mass attenuation coefficient is a mass density factor, the
connections between the interaction cross-section and the linear attenuation coefficient are the
The linear attenuation coefficient and mass attenuation coefficient were determined through an
experiment. In the experiment, measurements of beam intensity with and without an object layer
were used to determine the mass attenuation coefficient, which can be calculated using Equation
3.13 (Hubbell, 2006). The linear attenuation coefficient was then calculated by multiplying the
mass attenuation coefficient by the mass density using Equation 3.14. The fractional reduction of
beam intensity, -dI(l)/I, is proportional to the mass attenuation coefficient, µ/ρ, and the object
layer thickness, l (Equation 3.15). For a homogeneous object layer and mono-energetic beam, the
integration of Equation 3.15 reduces to the exponential attenuation law (Equation 3.16) for which
Equation 3.13 follows (Hubbell, 2006). Equation 3.16 calculates radiation attenuation.
𝐼0
𝜇/𝜌 = 𝑙 −1 𝑙𝑛 ( ) (3.13)
𝐼(𝑙)
57
𝐼0
𝜇 = 𝑙 −1 𝑙𝑛 ( )𝜌 (3.14)
𝐼(𝑙)
𝑑𝐼(𝑙) 𝜇
− = 𝑑𝑙 (3.15)
𝐼 𝜌
𝜇
𝐼(𝑡) = 𝐼0 𝑒𝑥𝑝(− 𝑙) (3.16)
𝜌
where µ/p is mass attenuation coefficient, usually in unit cm2 g-1, l is the matter
thickness of an object layer in unit of g cm-2, I0 is the intensity of the incident beam of
photons measured with the object layer removed from the beam, I(l) is the intensity of the
transmitted beam measured with the object layer interposed in the beam, µ is the linear
attenuation coefficient, and ρ is the density of the object layer.
Both the mass attenuation coefficient and the linear attenuation coefficient determined in the
experiment are limited by photon energy range. For a wide photon energy range, a theoretical
interaction cross-section has been compiled and maintained by various institutes, organisations, or
government agencies; for example, the National Bureau of Standards in the USA (name changed
in 1988 to the National Institute of Standards and Technology, NIST), the National Standard
Reference Data System in the USA, or the International Atomic Energy Agency (IAEA).
Theoretical interaction cross-section data for photon energy from 1 eV to 100 GeV were compiled
with hybrids of experimental data and theoretically calculated data. These data had been verified
The datasets of the mass attenuation coefficients and linear attenuation coefficients for Rayleigh
58
were retrieved from the XCOM data bank (Berger and Hubbell, 1987) which is currently
The differential cross-section is a fictitious cross-section that represents the probability of a photon
being scattered in a new direction. This new direction forms an angle (polar angle, θ) with this
photon’s direction before being scattered. The differential cross-section can be understood as the
ratio of the number of the detected scatter photons to the total number of photons in the beam.
In an experiment in which a detector is placed to detect the scatter photon at directions that form
an angle θ with the photon’s direction before the interaction (Figure 3.3), if the incident photon is
replaced by a mono-energetic beam, the detector will detect all photons scattered into these
directions. The number of detected photons over the total number of photons in the beam
measures of the angular distributions of scatter photons’ directions, they are essential for Monte
scattering and Compton scattering models. These theoretical models assumed that electrons are
free and not moving; however, electrons in matter are bound in atoms and are moving with
momentum. The effect of bound electrons (also known as the binding effect) on scatter photons’
directions was accounted for using the atomic form factor for Rayleigh scattering or the incoherent
59
The probability of the angular distribution of scatter photons’ directions was calculated using
Equation 3.17 for Rayleigh scattering (Hubbell et al., 1975) or Equation 3.18 for Compton
1+𝑐𝑜𝑠 2 𝜃
𝑃(𝜃) = [𝐹(𝑞, 𝑍)]2 (3.17)
2
𝐸 2 𝐸 𝐸
𝑃(𝜃) ≅ ( 𝐶 ) ( 𝐶 + − 𝑠𝑖𝑛2 𝜃) 𝑆(𝑞, 𝑍) (3.18)
𝐸 𝐸 𝐸 𝐶
𝐸
𝐸𝐶 = (3.19)
1+𝑘(1−𝑐𝑜𝑠𝜃)
where P(θ) is the probability of scattering at angle θ, E is the photon energy, F(q, Z)
and S(q, Z) are the atomic Form factor and incoherent scattering function, respectively, of
which values can be retrieved from NIST data bank, q is the momentum transfer variable
related to E and θ, and Z is the atomic (charge) number of the nucleus of the target atom,
𝐸
EC is the scatter photon energy, 𝑘 = , me is the electron mass, and c is a constant, the
𝑚𝑒 𝐶 2
speed of light in vacuum.
The atomic form factor only accounts for the effect of the bound electron which is undergoing the
interaction. It does not account for the effect of the other bound electrons of compounds on the
scatter photon (Morin, 1982, Poletti et al., 2002, Tartari et al., 2002, Poludniowski et al., 2009a,
Cunha et al., 2010). Investigations showed that this effect on the distribution of scatter radiation is
not negligible (Persliden and Carlsson, 1997, Cardoso et al., 2003, Poludniowski et al., 2009a,
Cardoso et al., 2009, Cunha et al., 2010) and it is referred to as the interference effect of compounds.
60
The interference effect can be accounted for using an interference function (Poludniowski et al.,
2009a). When the inference effect of compounds was accounted for, the probability of angular
1+𝑐𝑜𝑠 2 𝜃
𝑃(𝜃) = [𝐹(𝑞, 𝑍)]2 𝐺(𝑞, 𝑍) (3.20)
2
where P(θ) is the probability of scattering at angle θ, F(q, Z) and G(q, Z) are the atomic
form factor and interference function, respectively, q is the momentum transfer variable
related to E and θ, Z is the atomic (charge) number of the nucleus of the target atom, the
values of F(q, Z) can be retrieved from NIST data bank, and the values of G(q, Z) of some
materials can be retrieved from Tartari et al. (2002), Royle and Speller (1995), or Evans et
al. (1991).
A photon’s initial status was set to ‘primary’ when photons were generated at the focal spot.
A change in a photon’s energy was only made for Compton scattering. In Rayleigh scattering, a
photon was scattered without changing its energy. In photoelectric absorption, a photon was
absorbed in the matter. In Compton scattering, the energy change depended on the direction of the
scatter photon. The scatter photon energy was computed using the Compton (1923) equation
(Equation 3.19), which assumed that the electron was free and at rest (Bushberg et al., 2011).
Electrons in matter are orbiting their atoms which bind them. These electrons are moving with
interaction with a photon, can affect the scatter photon’s energy which is determined using the
Compton equation (Equation 3.19). The change in scatter photon energy ascribed to an electron’s
61
momentum is known as the Doppler broadening effect – the greater the Doppler broadening effect,
the lesser the photon energy lost. The Doppler broadening effect only happens in Compton
scattering.
In the above discussions, random numbers were essential and were extensively used in photon
transportation. The generation of random numbers and number sampling methods are discussed in
In Monte Carlo simulation, many decisions for events to occur depend on random numbers and
criteria (such as bigger than, equal to, or smaller than predefined values). The frequency of an
event’s occurrence is often dependent on the frequency of a range of number being sampled.
Sampling numbers for the occurrences of events is a vital component of Monte Carlo simulation.
numbers are often generated using a deterministic algorithm, which means these random numbers
are generated in a predictable sequence. A deterministic algorithm that generates random numbers
can be written as a software function, which is commonly referred to as a random number generator.
When a random number generator is executed, numbers are generated in a sequence. For example,
when the nth time a random number generator is executed, the nth number in the sequence is
returned. When the last number in the sequence is returned, the next time the generator is executed,
the generator will return the 1st number in the sequence. Therefore, this sequence is periodic.
62
A random number generator constructed using Equation 3.21 produces a sequence with a period
of the order of 109 (Press and Teukolsky, 1992). This generator is like a random number generator
used in Matlab versions older than 5 (The MathWorks Inc., Natick, Massachusetts). In Matlab
version 7.4 and later, a new uniform random number generator uses an algorithm known as the
Mersenne Twister (M. Matsumoto and Nishimura, 1997). This generator has an incredible period
of length 219937 − 1. This generator was used in the Monte Carlo simulation in this project.
Random numbers generated using a deterministic algorithm are known as pseudo-random numbers
because they are not random but have excellent random properties. The advantage of using pseudo-
random numbers is that they are efficiently generated with tens of gigabit per second rates, which
is only limited by computing speed and resources (Press and Teukolsky, 1992).
True random numbers can be generated, such as using a hardware that generates random numbers
from a physical process. Hardware random number generators generally generate a limited number
of random numbers per second. Some fast hardware random number generators can generate
random numbers with multi-gigabit per second rates (Marandi et al., 2012).
Sampling methods determine the frequency of a range of numbers being sampled. The most
common sampling method is uniform sampling, in which any number has an equal probability of
63
being sampled. In Monte Carlo simulation, many probabilities are not uniformly distributed.
Different distributions can be sampled more efficiently using different sampling methods.
In this project, the probability of a photon travelling a distance without any interaction with matter
was calculated using Equation 3.22, which was deduced from Equation 3.16. The probability
function (P(l)) (Equation 3.22) has a constant probability for any constant path length (l) in a
uniform media where µ/p is a constant. This probability function represents a uniform distribution.
As a free path length (l) is associated with the cumulative probability of interactions from path
length zero to l, the uniformly distributed probability function (P(l)) cannot be correctly sampled
by uniformly sampling the path length, which has a one-to-one relationship with the cumulative
probability of P(l). Equation 3.22 was rewritten as Equation 3.23, in which the random number ξ,
was uniformly distributed in the interval (0, 1). The free path length then can be sampled by
𝜇
𝜉 = 𝑃(𝑙) = 𝑒𝑥𝑝(− 𝑙) (3.22)
𝜌
1
𝑙= − 𝜇 𝑙𝑛(1 − 𝜉) (3.23)
−
𝜌
where ξ or P(l) is the probability, µ/p is mass attenuation coefficient, and l is the free
path length.
64
3.3.3.2.2. Discrete indexed sampling
A discrete indexed sampling method uniformly samples the indices of a list/table, and each time
the sampled index is used to retrieve the value at this index. Any probability distribution (either
discrete or continuous) of random events can be converted into a list or table, in which each of
these events is assigned with a range of indices. A sequence of these random events is then
where x is a random variable. F(x) is the frequency of x. Despite x and F(x) being continuous or
discrete, they can be divided into equal discrete intervals. The more the intervals, the better the
representation of this frequency; in other words, the more the intervals, the higher the accuracy.
Figure 3.4(b) illustrates the area under the curve divided into small rectangular areas that are
distinguished by grid lines. On the x-axis, a discrete interval of x determines a column. The first
column, x1, starts at the smallest x. The last column, xmax, ends at the largest x. On the y-axis, a
discrete interval of F(x) determines a row. The first row starts at zero; the last row ends at F(x).
The total number of the small rectangular areas under the curve is the sampling resolution; the
higher the resolution, the better the representation of F(x). A count of these areas can be established
starting at the first column from top to bottom, then at the second column from top to bottom, and
so on, until reaching the last row in the last column. A list of the probability F(x) is shown in Table
3.1.
65
Table 3.1. Indices and variable values
Index 1 2 3 4 5 6 7 8 9 10 … Countmax
6 6 (b)
5 (a) 5
4 4
F(x)
F(x)
3 3
2 2
1 1
0 0
x
x
Figure 3.4. Indexed discrete sampling. (a) Plot of the non-uniform probability function F(x); (b)
Discrete areas in small rectangles under the curve.
A reject-accept sampling method uniformly samples a random number and tests the number under
a condition. If the condition is satisfied, the number is accepted; otherwise, the number is rejected.
If the number is rejected, this process continues until a random number is accepted.
Many non-uniform probability distributions can be sampled using reject-accept sampling. In many
situations, reject-accept sampling can be replaced by discrete indexed sampling and appropriate
results can still be achieved. The disadvantage of reject-accept sampling is that this sampling may
be very inefficient. The disadvantage of the discrete indexed sampling method can be that a very
66
3.4. Implementation of interactions between photons and matter
The interactions between photons and matter were implemented using Rayleigh scattering,
Compton scattering, photoelectrical absorption, and pair production (in nuclear field and in
electron field) models. A pair production can only be possible if photon energy is at least twice the
energy of an electron mass; that is, approximately 512 × 2 kilo-electron-volt (keV). In x-ray
imaging, photons have energy approximately 10–150 keV. Therefore, the relevant interactions
between these photons and matter are Rayleigh scattering, Compton scattering, and photoelectrical
calculated as the ratios of the cross-section of each interaction type to the total interaction cross-
In Rayleigh scattering, a photon interacts with the electrons of an atom as a whole, and after the
interaction the photon changes its direction and does not lose its energy. Bushberg et al. (2011)
suggested that the photon interacts with and excites the whole atom through the energy from the
photon’s electric field, which causes all electrons of the atom to oscillate in phase. The oscillation
of these electrons immediately radiates a photon (scatter photon) of the same energy in a slightly
different direction. According to Hendee and Ritenour (2002), a photon is assumed to interact with
the atom’s electrons as a whole, and the photon is deflected or scattered with a negligible loss of
energy. Generally, Rayleigh scattering is implemented as such that after the interaction the photon
changes its direction without losing energy (Chan and Doi, 1982, Boone and Cooper III, 2000,
Nikolopoulos et al., 2007, Cunha et al., 2010, Khodajou-Chokami and Sohrabpour, 2015).
67
The direction of a scatter photon was determined using two angles: a polar angle θ and an azimuth
angle ω (Figure 3.3). The polar angle was determined per probability distribution function given
in Equation 3.17. Previous discussion (Section 3.3) had shown that this distribution does not
account for the interference effect of compounds. When the interference effect was accounted for,
the polar angle was determined per probability distribution function in Equation 3.20. The azimuth
angle was uniformly sampled in (0, 2π] with angle units in radians. The polar angle, the azimuth
angle, and the photon direction cosines before the interaction determined the direction of the scatter
photon. The scatter photon direction was then calculated using the equations given in Section 3.3.1.
In Compton scattering, a photon interacts with a bound electron and causes the electron to be
emitted from its atom. The photon loses some energy and is scattered to become a scatter photon.
The scatter photon’s direction was determined using two random angels: a polar angle θ and an
azimuth angle ω (Figure 3.3). The polar angle was sampled from a probability distribution function
given in Equation 3.18. Like in Rayleigh scattering, the azimuth angle was uniformly sampled in
(0, 2π] with angle units in radian. The polar angle, the azimuth angle, and the photon direction
cosines before the interaction determined the direction of the scatter photon. This direction was
The energy of the scatter photon was affected by not only the photon scattering direction but also
by the electron momentum before the interaction. The effect of the electron’s momentum on the
scatter photon energy is known as the Doppler broadening effect. Section 3.3.2 discussed that the
Doppler broadening effect may cause the energy of the scatter photon at a direction to change and
68
that for a fixed scattering angle, the energy of the scatter photon of a mono-energetic beam may
In x-ray imaging simulation, the Doppler broadening effect has been found to have negligible
effects on scatter radiation (Cunha et al., 2010) and can be accounted for by using an appropriate
distribution of energy broadening/changing. This project did not account for the Doppler
broadening effect. The scatter photon energy was computed using Equation 3.19.
Emitted electrons in tissues have low kinetic energy, less than a few keV, because of photons’
energy (<150 keV) as well as the low atomic number materials of tissues. An electron only travels
in soft tissues for a very short distance (<1 mm). Any secondary electrons or photons excited or
emitted by this electron also only survived a very short distance (<1 mm). The tracking of the
emitted electrons was neglected, and their kinetic energy was locally deposited at the interaction
site (Chan and Doi, 1982, Boone and Cooper III, 2000, Nikolopoulos et al., 2007, Cunha et al.,
photon interacts with a bound electron. The photon loses all of its energy and causes the electron
to be emitted with kinetic energy equal to the photon energy minus the electron’s binding energy.
Photoelectric absorption only occurs if the photon energy is greater than the electron binding
energy. The tracking of emitted electrons was neglected and their kinetic energy was locally
deposited at the interaction site (Chan and Doi, 1982, Boone and Cooper III, 2000, Nikolopoulos
69
3.5. Radiation dosimetry in the Monte Carlo simulation
Radiation dosimetry is the measure of radiation. Measuring the amount of radiation is another
essential component of Monte Carlo simulation and is made in derived units from seven basic
units. The seven basic units are provided by the International System of Units (SI) for science and
technology and are metre for length, kilogram for mass, second for time, ampere for electric current,
kelvin for temperature, candela for luminous intensity and mole for substance. The detailed
Radiation Units and Measurements (2011). Several units that are commonly used in medical x-ray
Photon fluence is the number of photons or particles passing through a unit cross-section area,
typically expressed in units of cm-2. The usual symbol Φ is given to photon fluence and calculated
using Equation 3.24 (International Commission on Radiation Units and Measurements, 2011).
Photon flux is the rate of photon fluence and is calculated using Equation 3.25. Photon flux is
useful for measuring photons over long periods (tens of seconds), such as in fluoroscopy. The unit
for photon flux is usually cm-2 s-1 (International Commission on Radiation Units and
Measurements, 2011).
Energy fluence is the amount of energy passing through a unit cross-section area. For a mono-
energetic beam of photons, energy fluence (Ψ) is the product of photon fluence (Φ) and the energy
per photon (E). The typical unit for energy fluence is J m-2. Energy fluence of a mono-energetic
beam is calculated using Equation 3.26 (International Commission on Radiation Units and
Measurements, 2011).
70
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝ℎ𝑜𝑡𝑜𝑛𝑠
𝛷= (3.24)
𝑎𝑟𝑒𝑎
𝛷
𝐹𝑙𝑢𝑥 = (3.25)
𝑡𝑖𝑚𝑒
𝛹 = 𝛷𝐸 (3.26)
where Φ is photon flux, Ψ is the energy fluence, and E is the photon energy.
In an interaction between a photon and a charged particle, the photon may lose a fraction of its
energy to this particle as the particle’s kinetical energy plus, if any, its binding energy. The mass
energy transfer coefficient is the total mass attenuation coefficient multiplied by this fraction. The
mass energy transfer coefficient is given as μtr/ρ (International Commission on Radiation Units
The mass energy absorption coefficient accounts for the fraction of the mass energy transfer
coefficient that gives rise to the total energy locally deposited at the interaction site (International
Commission on Radiation Units and Measurements, 2011). After a charged particle gains kinetic
energy, its kinetic energy is deposited in matter through the subsequent excitation or emission of
other charged particles. Eventually, the kinetic energy of all charged particles is absorbed in the
matter. However, in interactions in which scatter photons are produced, the scatter photon energy
may not be deposited in the matter. Therefore, the mass energy absorption coefficient is always
71
With high atomic number materials, such as lead, several tens of keV of electrons emitted in
caused by an electric field of another charged particle (Bushberg et al., 2011). Some of the higher
energy (several tens keV) bremsstrahlung radiation may escape the interaction site. Therefore, the
mass energy absorption coefficient is generally smaller than the mass energy transfer coefficient
Approximately 96% of the human body’s matter is composed of low atomic number elements
(oxygen 65.0%, carbon 18.5%, hydrogen 9.5%, and nitrogen 3.2%) (Tortora and Derrickson,
2008). In x-ray imaging, photons interact with body matter mainly through Compton scattering.
Electrons emitted by these photons in Compton scattering have low kinetic energy (less than 10
keV), and these electrons produce low energy bremsstrahlung radiation which excites or emits
other electrons with low kinetic energy. The kinetic energy of all these electrons is deposited within
less than one millimetre of the interaction site. For body tissues, the mass energy absorption
KERMA stands for kinetic energy released per unit mass and is the kinetic energy transferred to
charged particles by uncharged ionising radiation per unit mass. The SI unit for KERMA is J kg-1
with the special name of the Gray (Gy) or milliGray (mGy) (1 Gy = 1000 mGy = 1 J kg-1). KERMA
can be calculated using Equation 3.27 as the product of the mass energy transfer coefficient and
72
𝜇𝑡𝑟
𝐾 = 𝛹( ) (3.27)
𝜌
where K is KERMA, μtr/ρ is the mass energy transfer coefficient, and Ψ is energy
fluence.
When uncharged ionisation radiation, such as x-ray photons, passes through a matter, the
deposition of radiation energy in the matter happens in two steps. In the first step, energy of the
x-ray imaging, photon energy is less than 150 keV, and the photon energy is transferred to electrons
through the relevant photoelectric absorption or Compton scattering. In the second step, the
charged particles deposit their energy in the matter by exciting and ionising other electrons.
Eventually, all the kinetic energy of electrons is deposited near the interaction site because of the
very short travelling distances (<1 mm) of these electrons and, if any, bremsstrahlung radiation
(less than a few keV) originating from acceleration/deceleration of these electrons (International
The absorbed dose is the energy deposited through ionisation in unit mass of irradiated material.
The absorbed dose is calculated using Equation 3.28 and the SI unit is J kg-1 or Gy. The difference
between the absorbed dose and the KERMA is that KERMA is only defined for uncharged ionising
radiation, but the absorbed dose is defined for both charged and uncharged ionising radiation. A
historical unit, the rad, is often used for the absorbed dose and one rad is equal to 0.01 J kg-1 (1 rad
= 0.01 J kg-1 = 10 mGy) (International Commission on Radiation Units and Measurements, 2011).
73
𝐸
𝐷= (3.28)
𝑚
where D is the absorbed dose, E is the total energy departed, and m is the mass.
The absorbed dose is one of the basic dosimetric quantities. Different medical x-ray imaging
technologies may use specific dosimetric quantities. Computed tomography (CT) often uses the
CT dose index, the dose-length product, or multiple scan average dose. Fluoroscopy uses the entry
skin dose or KERMA area product (KAP). Mammography uses the mean/average glandular dose
Radiation exposure is the amount of electrical charge that is produced by ionising radiation per
unit mass of air. Radiation exposure is calculated using Equation 3.29. The historical unit of
radiation exposure is the Roentgen, and the SI unit for radiation exposure is coulombs per kg (1
Roentgen = 2.58 x 10–4 C kg-1) (International Commission on Radiation Units and Measurements,
2011).
𝑄
𝑋= (3.29)
𝑚
where X is the radiation exposure, Q is the total amount of charges of one sign, and m
is the mass.
In medical diagnostic imaging, parameters given to an x-ray generator to generate photons are
often in milliampere second (mAs) and kilovoltage peak (kVp). The output intensity of an x-ray
generator can be measured and expressed as a radiation exposure per unit mAs under specific
74
operating conditions. An example of radiation exposure is 10 mR/mAs at 70 kVp at a source-
Radiation exposure can be directly measured with air-filled radiation detectors (air-chamber
dosimeters). The effective atomic numbers of air and soft tissue are approximately the same.
Radiation exposure is nearly proportional to the absorbed dose in soft tissue over the range of
photon energies used in diagnostic radiology (International Commission on Radiation Units and
Measurements, 2011).
Radiation exposure can be calculated from an average energy deposition per ion pair in air, which
is approximately constant at 33.85 eV/ion pair or 33.85 J/C. The relationship between the absorbed
dose, radiation exposure, and air KERMA is given by Equation 3.30. For example, one Roentgen
of radiation exposure results in approximately 8.73 mGy absorbed dose in air (International
𝐷 𝐸
𝑊= = (3.30)
𝑋 𝑄
where W is the average energy deposition per ion pair in air, D is the absorbed dose, X
is the radiation exposure, E is the total energy deposition in matter, and Q is the total amount
of charges of one sign.
In simulation, the details of phantoms are constructed as dummy anatomies that facilitate the study
of imaging systems, imaging procedures, or patient radiation dose. Phantoms can be constructed
75
using polynomial boundaries of materials or using Hounsfield units of volume data from CT.
Polynomial boundaries are best for uniform phantoms or phantoms with several types of uniform
materials. Hounsfield units of volume data from CT is best for heterogeneous phantoms mixed
with many materials. The heterogeneous phantom constructed in Badal and Badano (2009) is an
Summary
This chapter discussed details of the simulation of photon transportation in x-ray imaging. These
details were used to develop our Monte Carlo simulation code system. A simulation procedure for
photon transportation was developed, which followed a photon’s track from the x-ray focal spot
to the image receptor. Details of photon transportation were then examined, including: (1) changes
in a photon’s location, direction, and/or energy; (2) random number generation and sampling
methods; (3) relevant interactions between photons and matter; and (4) radiation dosimetry and
The methods for measuring the changes in a photon’s location, direction, and/or energy were
discussed and established. Photon location change was determined using a free path that was
determined using a random number and the linear attenuation coefficient. Photon direction change
was determined using the photon direction and two random angles, an azimuth angle uniformly
sampled in (0, 2π] and a polar angle sampled in (0, π] with all angle units in radian. The polar
angle depended on the atomic form factor for Rayleigh scattering or on the incoherent scattering
function for Compton scattering. In Rayleigh scattering, the interference effect of compounds on
the polar angle could be accounted for by using an interference function. Photon energy change
was calculated using the Compton equation. The Doppler effect on the scatter photons’ energy
76
broadening was ignored. This effect could be implemented using a probability distribution of the
A random number generator and sampling methods were analysed for simulating photon
transportation. Uniform sampling was generally used in the simulation. An indexed sampling
method was introduced to sample non-uniformly distributed probability for improved sampling
efficiency. This indexed sampling method gave each probable event the proportional number of
indices to the event’s probability. When this method was executed, an index was uniformly
sampled from the indices. The sampled index was used to determine which event will happen. A
reject-accept sampling method was also discussed for sampling non-uniformly distributed
probability. In this method, a random number was uniformly sampled and tested under a condition.
If the condition was satisfied, the number was accepted; otherwise, the number was rejected. If the
number was rejected, this process continued until a random number was accepted
The relevant interactions (Rayleigh scattering, Compton scattering, and photoelectric absorption)
between photons and matter were implemented in our Monte Carlo simulation code system.
Rayleigh scattering was implemented with only change in photon direction. Compton scattering
was implemented with changes in photon direction and energy. Photoelectric absorption was
implemented as photon energy was completely deposited at the interaction site. In all the
interactions, the kinetic energy of emitted electrons was locally deposited in the matter. The
In addition, radiation dosimetry was reviewed and its implementation was discussed. Two virtual
phantom construction methods were also discussed: one used the polynomial boundaries of
materials and the other used the Hounsfield units of volume data from CT.
77
The virtual phantom construction methods, however, were not suitable for constructing a grid for
the simulation of photon transportation in the grid materials. The next chapter will solve the
78
Chapter 4: New method for determining radiation
transmission in grids
This chapter addresses photon transportation in grid materials. It reviews contemporary methods
for radiation transmission in grids and discusses these methods’ limitations. It then develops and
validates a new method, to overcome these limitations. The article related to the work of this
Basic grid performance parameters of transmission of primary, scatter, and total radiation are
discussed in Chapter 2. For physical grids, these transmissions are determined using the amounts
of radiation measured with or without grids (Bonenkamp and Hondius Boldingh, 1959a, Strid,
1976, Fetterly and Schueler, 2007, Fetterly and Schueler, 2009, Salvagnini et al., 2012).
A grid’s performance can be assessed through its design evaluation by ignoring manufacturing
defects, if any. Grid designs can be evaluated through Monte Carlo simulation, which are used to
determine grid designs’ radiation transmissions using analytical methods (Dance and Day, 1984,
Dance et al., 2000, Malusek et al., 2003, Ullman et al., 2006, Malusek et al., 2008, Star-Lack et
al., 2009, Lazos and Williamson, 2010, Sun and Star-Lack, 2010, Zbijewski et al., 2012, Sisniega
Several analytic methods (Strid, 1976, Kalender, 1982, Bernstein et al., 1983, Day and Dance,
1983) have been developed to determine grid designs’ radiation transmissions. First, these methods
79
assumed that both primary radiation and scatter radiation were uniformly distributed at the grid
entry surface. Generally, primary radiation at the grid entry surface is approximately uniformly
distributed. However, it is well known that scatter radiation is non-uniformly distributed (Boone
et al., 2000, Johns and Yaffe, 1982). Second, these methods determined the radiation mean
transmission in each grid unit, which consists of a grid strip and an adjacent interspace region.
These methods, therefore, eliminate the details of grid lines resulted from stationary grids. Grid
lines in an x-ray image appear as white strips, which may interfere with the interpretation of the
Strid (1976) proposed an analytical method to calculate the transmission of scatter radiation, of
which angular distribution was determined using sample measurements in different directions.
These measurements were obtained using an ionisation chamber detector, which was equipped
with a long, slender, cylindrical collimator. The collimator was rotated to allow the scatter
Strid (1976) assumed scatter radiation was distributed symmetrically about the centre of a parallel
grid and designated four exclusive regions for scatter radiation entering each grid unit. In the first
three regions, exact transmissions were determined. The transmission in the fourth region was
extrapolated by ignoring the radiation that would travel to at least another grid unit. For focused
grids, Strid (1976) determined an approximate transmission of scatter radiation by treating focused
Kalender (1982) determined radiation transmission by using a weighting factor and the total
attenuation of grid materials. The determination of the weighting factor depends on radiation path
lengths in the grid materials. Day and Dance (1983) suggested that this weighting factor could be
80
determined from the scatter-radiation-energy-angle spectrum, which can be either determined
Bernstein et al. (1983) extended the Strid (1976) method to obtain better approximation of
radiation transmissions in focused grids. Bernstein et al. (1983) designated three distinct regions
instead of the four distinct regions of Strid (1976). Furthermore, Bernstein et al. (1983) determined
the photon path lengths in focused grid material by increasing the photon’s polar angle by another
angle, which was equal to the angle of the strip relative to the grid normal (this grid normal is
parallel to the z-axis in Figure 4.1). However, Day and Dance (1983) found that Bernstein et al.
(1983) did not determine the correct radiation path lengths in focused grid materials.
Following these issues with the methods of Strid (1976), Kalender (1982), and Bernstein et al.
(1983), Day and Dance (1983) proposed a new method to determine the radiation transmissions in
oblique grids, which had their strips in parallel. Day and Dance (1983) also assumed that radiation
was uniformly distributed at the grid’s entry surface. The Day and Dance method precisely
determined the grid-unit mean transmissions of radiation in parallel grids and extrapolated
radiation mean transmissions in focused grids by replacing strip angles with new angles. These
new angles were calculated using two values: (1) the x-coordinate (Figure 4.1) of the radiation
entry points on the grid and (2) the grid focal distance. The Day and Dance method is used in the
Monte Carlo simulation to determine radiation transmissions in grid materials (Dance and Day,
1984, Dance et al., 2000, Malusek et al., 2003, Ullman et al., 2006, Malusek et al., 2008, Star-Lack
et al., 2009, Lazos and Williamson, 2010, Sun and Star-Lack, 2010, Zbijewski et al., 2012,
81
Figure 4.1. Strip alignments of focused grids showing the geometrical relationships of the strip
thickness (d), interspace (D), and strip height (h). A Cartesian system is defined by the x-, y-,
and z-axes perpendicular and parallel to the strip direction, and perpendicular to the plane of
the grid, respectively. The centre of the grid is located at C(0, 0, f0) where f0 is the grid focal
distance. The line segment AAexit illustrates a photon trajectory passing through the grid.
Three problems for all the above methods are: (1) inaccurate assumption of scatter radiation
which does not preserve the details of grid lines resulting from stationary grids; (3) extrapolation
of radiation transmissions for focused grids. In the following sections, a new method will be
A photon has a probability for travelling through grid materials without interacting with them. In
this case, the photon may have travelled in the grid’s strip material and/or its interspace material,
and the photon’s path lengths in each of these materials can be determined. This probability can
then be calculated using these lengths and these materials’ linear attenuation coefficients. In the
82
Contemporary methods assume that photons at the grid entry surface are uniformly distributed.
This new method is not constrained by the distribution of photons at the grid entry surface. In
addition, this and other methods ignore scatter radiation and fluorescent radiation arising from the
grid materials. because the amounts of those reaching the image receptor are negligible (Strid,
1976, Kalender, 1982, Bernstein et al., 1983, Day and Dance, 1983).
Photon path lengths in grid materials were determined using a geometrical constraint. In a grid, a
strip surface is the interface surface between the strip and interspace materials (Figure 4.1). This
constraint was defined by all points on the same strip surface; that is, a constant x-coordinate for
For all types of grids, a three-dimensional Cartesian system was defined by X-, Y-, and Z-axes
perpendicular and parallel to the strip direction, and perpendicular to the plane of the grid,
respectively. The centre of grid entry plane was located at C(0, 0, f0). The f0 is the focal distance
for a focused grid or the distance from the focal spot to the grid entry plane of a parallel grid
(Figure 4.1). The x-ray focal spot was located at the origin O(0, 0, 0). When an arbitrary photon
entered the grid at point A(x, y, z) on the grid’s entry plane and moved in the direction given by
the direction cosines (Δx, Δy, Δz) without any interaction, this photon might intersect the grid’s
strip surfaces and exit the grid at point Aexit(xexit, yexit, zexit) on the grid’s exit plane (Figure 4.1).
Similar to the Day and Dance (1983) formula, the transmission of the photon along path AAexit was
83
where T is the transmission of a photon, ls and li are this photon’s path lengths in the
strip and interspace region, respectively, and us and ui are the linear attenuation coefficients
of the strip and interspace materials, respectively.
If this photon intersected any strip surfaces, the coordinate of the ith intersection point was
expressed as Mi(x+aiΔx, y+aiΔy, z+aiΔz), where ai was a variable. The value of the variable ai was
solved in the following sections. The coordinates of point A, point Aexit, and, if any, all Mi were
used to calculate ls and li. The linear attenuation coefficients (Section 3.5) can be retrieved from
the XCOM data bank (Berger and Hubbell, 1987) currently maintained by National Institute of
The spatial relationships between point A, point Aexit, grid strip, and grid interspace determined the
expression of the variable ai. These relationships were exhaustively classified by ten mutually
exclusive conditions (Table 4.1). An illustration of these relationships is shown in Figure 4.2, and
these conditions’ criteria are detailed in Table 4.2. The solution of variable ai is detailed in Section
84
Figure 4.2. Photon’s trajectories through the grid materials: C1, C2, C5, C6, and C7 illustrate
trajectories from an interspace region in the grid entry surface to a region in the grid exit surface;
C3, C4, C8, C9, and C10 illustrate trajectories from a strip region in the grid entry surface to a
region in the grid exit surface.
Table 4.1. Ten categories (Ci) of location relationships of points A(x, y, z) and A’(x’,
y’, z’). The conditions are restricted for x ≥ 0. For x ≤ 0, as it is mirror symmetric to x
≥ 0, results can be determined by applying a ‘minus’ sign to the x-coordinates.
C1 (x > x’) C2 (x > x’) C3 (x > x’) C4 (x > x’) C5 (x = x’)
A(x, y, z) i i s s i
C6 (x < x’) C7 (x < x’) C8 (x < x’) C9 (x < x’) C10 (x = x’)
A(x, y, z) i i s s s
85
Table 4.2. The determination criteria of the ten categories (Ci) of the location
relationships of points A(x, y, z) and A’(x’, y’, z’). The conditions are restricted for x ≥
0. For x ≤ 0, as it is mirror symmetric to x ≥ 0, results can be determined by applying
a ‘minus’ sign to the x-coordinates.
Categories criteria of categories
where r is the mantissa of (𝑥 + 𝐷/2)/(𝐷 + 𝑑); r’ is the mantissa of (𝑥′ + 𝐷/2)/(𝐷 + 𝑑).
In Figure 4.1, the photon entered the grid’s entry surface at A(x, y, z) and exited the grid at Aexit(xexit,
yexit, zexit). The straight line between the origin O(0, 0, 0) and Aexit(xexit, yexit, zexit) intersected the
grid’s entry plane at point A’(x’, y’, z’). Ten possible relationships between point A and point A’,
called categories (Ci, i = 1, 2,…, 10) (Table 4.2), were determined with two mantissas. A mantissa
is the positive fractional part of real number; for example, for x = 3.1415, the mantissa is 0.1415.
𝐷
𝑥+
One of the two mantissas was equal to the mantissa of 𝐷+𝑑, where x was the x-coordinate of point
2
𝐷
𝑥′+
A. The other mantissa was equal to the mantissa of , where 𝑥′ was the x-coordinate of point A’.
2
𝐷+𝑑
86
Points A and A’ might be located in different grid units. The difference between the different grid
𝐷 𝐷
𝑥+ 𝑥′+
𝑁 = |𝐼𝑁𝑇 (𝑑+𝐷2 ) − 𝐼𝑁𝑇 ( 𝑑+𝐷2 )| (4.2)
where N is the number of the different grid units, |a| is the absolute value of ‘a’, INT(a)
is the largest integer equal to or smaller than ‘a’, d is the strip thickness, D is the interspace
distance, x is the x-coordinate of point A(x, y, z), and 𝑥′ is the x-coordinate of point A’(x’,
y’, z’).
All points on the same strip surface had the ratios of their x-coordinates to their z-coordinates
equal; that is, x/z = constant = T, where T was the common ratio on a strip surface. Ti denoted the
common ratio of the strip surface when the photon on the ith intersected a strip surface. Only
photons with x-coordinates equal to or greater than zero (x ≥ 0) were discussed. Solutions for
photons when x < 0 were obtained by a mirror-symmetrical operation about the y-z plane. Photons
with a directional component equal to or smaller than zero (Δz ≤ 0) were excluded from the
discussion. These photons were not directly moving towards the image receptor. In this project,
these photons were neglected because the fraction of them reaching the image receptor is trivial.
In category C1, N was a whole number greater than zero. If N was equal to 1, only one intersection
point M1(x+a1*Δx, y+a1*Δy, z + a1*Δz) existed from point A to point Aexit. For category C1 these
results hold:
𝑥−𝑟
𝑇1 = (4.3)
𝑓0
𝐷
𝑥+
where r is the mantissa of 𝐷+𝑑. 2
87
𝑥+𝑎1 ×∆𝑥
𝑇1 = (4.4)
𝑧+𝑎1 ×∆𝑧
(𝑇1 × ∆𝑧 − ∆𝑥) × 𝑎1 = 𝑥 − 𝑇1 × 𝑧
(4.5)
𝑥
𝑇1 ≠ (4.6)
𝑧
𝑇1 × ∆𝑧 − ∆𝑥 ≠ 0 (4.7)
1 𝑥−𝑇 ×𝑧
𝑎1 = 𝑇 ×∆𝑧−∆𝑥 (4.8)
1
The straight lines AM1 and M1Aexit were the trajectories that the photon might travel in the
interspace material and strip material, respectively, where linear propagation of the photon within
the grid materials was assumed. From the definition of the cosine directions, the value of a1, always
greater than 0, was equal to the length of AM1. The value of a1 simplified the calculations of the
If N was equal to 2, the first intersection M1(x+a1*Δx, y+a1*Δy, z + a1*Δz) was still valid. Two
𝑥−𝑟−𝑑
𝑇2 = (4.9)
𝑓0
𝑥−𝑟−𝑑−𝐷
𝑇3 = (4.10)
𝑓0
88
𝐷
𝑥+
where r is the mantissa of 𝐷+𝑑. 2
𝑥−𝑇 ×𝑧
2
𝑎2 = 𝑇 ×∆𝑧−∆𝑥 (4.11)
2
𝑥−𝑇 ×𝑧
3
𝑎3 = 𝑇 ×∆𝑧−∆𝑥 (4.12)
3
The values of a2 and a3 were equal to the lengths of AM2 and AM3, respectively.
For photon paths in category C1, N might be greater than 3 (indeed, without any interaction with
grid materials, N was limited to the number of grid strips). As N increased, the expressions of the
intersection points followed a general pattern. This pattern led to a general expression for photons
in category C1.
Similar to C1, general expressions for all the categories were obtained:
ℎ
𝑥𝑒𝑥𝑖𝑡 = 𝑥 + ∆𝑥 × ∆𝑧 (4.13)
ℎ
𝑦𝑒𝑥𝑖𝑡 = 𝑦 + ∆𝑦 × ∆𝑧 (4.14)
ℎ
𝑧𝑒𝑥𝑖𝑡 = 𝑧 + ∆𝑧 × ∆𝑧 (4.15)
𝑓
𝑥 ′ = 𝑥𝑒𝑥𝑖𝑡 × ℎ+𝑓
0
(4.16)
0
𝑓
𝑦 ′ = 𝑦𝑒𝑥𝑖𝑡 × ℎ+𝑓
0
(4.17)
0
𝑓
𝑧 ′ = 𝑧𝑒𝑥𝑖𝑡 × ℎ+𝑓
0
(4.18)
0
𝑥𝑖 = 𝑥 + 𝑎𝑖 ∆𝑥 (4.19)
𝑦𝑖 = 𝑦 + 𝑎𝑖 ∆𝑦 (4.20)
𝑧𝑖 = 𝑧 + 𝑎𝑖 ∆𝑧 (4.21)
𝑥−𝑧×𝑇
𝑖
𝑎𝑖 = 𝑇 ×∆𝑧−∆𝑥 (4.22)
𝑖
89
𝐷
𝑥+
Given that r is the mantissa of 𝐷+𝑑: 2
𝑖 𝑖−1
𝑥−𝑟−𝑑×𝐼𝑁𝑇( )−𝐷×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 − 1 (4.23)
𝑓0
For category C2: if N = 0, the photons would only travel in interspace material. For N > 0,
𝑖 𝑖−1
𝑥−𝑟−𝑑×𝐼𝑁𝑇( )−𝐷×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 (4.24)
𝑓0
For category C3: if N = 0, the photons would only travel in strip material. For N > 0,
𝑖−1 𝑖
[ 𝑥−𝑟+𝐷 ]−𝑑×𝐼𝑁𝑇( )−𝐷×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 (4.25)
𝑓0
𝑖 𝑖−1
[ 𝑥−r+𝐷]−𝐷×INT( )−𝑑×INT( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 + 1 (4.26)
𝑓0
For category C5, the photon would only travel in interspace material.
𝑖 𝑖−1
[ 𝑥−𝑟+𝐷]+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 + 1 (4.27)
𝑓0
For category C7: if N = 0, the photon would only travel in interspace material. For N > 0,
𝑖 𝑖−1
[ 𝑥−𝑟+𝐷]+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 (4.28)
𝑓0
For category C8: if N = 0, the photon would only travel in strip material. For N > 0,
𝑖 𝑖−1
[ 𝑥−𝑟+(𝑑+𝐷)]+𝐷×𝐼𝑁𝑇( )+𝑑×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 (4.29)
𝑓0
90
𝑖 𝑖−1
[ 𝑥−𝑟+(𝑑+𝐷)]+𝐷×𝐼𝑁𝑇( )+𝑑×𝐼𝑁𝑇( )
𝑇𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 − 1 (4.30)
𝑓0
For category C10, the photon would only travel in strip material.
Parallel grids were made of parallel strips where points on the same strip surface had their x-
coordinates equal in the coordinate system in Figure 4.1. A constant x-coordinate was used to
determine intersections on each grid strip surface. The relationships given in Tables 4.1 and 4.2
were valid for parallel grids, except that point A’ was replaced by point Aexit (Figure 4.1). The
general expressions, which are different from that for focused grids, were given for Δx ≠ 0 (for Δx
𝑥 ′ = 𝑥𝑒𝑥𝑖𝑡 (4.31)
𝑦 ′ = 𝑦𝑒𝑥𝑖𝑡 (4.32)
𝑧 ′ = 𝑧𝑒𝑥𝑖𝑡 (4.33)
𝑖 𝑖−1
𝑟+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
𝑎𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 − 1 (4.34)
|∆𝑥|
For category C2: if N = 0, the photon would only travel in interspace material. For N > 0,
𝑖 𝑖−1
𝑟+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
𝑎𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 (4.35)
|∆𝑥|
For category C3: If N = 0, the photon would only travel in strip material. For N > 0,
𝑖−1 𝑖
( 𝑟−𝐷)+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
2 2
𝑎𝑖 = 𝑖 ∈ 1, 2, 3, … , 2𝑁
|∆𝑥|
(4.36)
91
𝑖 𝑖−1
( 𝑟−𝐷)+𝐷×𝐼𝑁𝑇( )+𝑑×𝐼𝑁𝑇( )
2 2
𝑎𝑖 = 𝑖 ∈ 1, 2, 3, … , 2𝑁 + 1
|∆𝑥|
(4.37)
For category C5, the photon would only travel in interspace material.
𝑖 𝑖−1
( 𝐷−𝑟)+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
2 2
𝑎𝑖 = 𝑖 ∈ 1, 2, 3, … , 2𝑁 + 1
∆𝑥
(4.38)
For category C7: if N = 0, the photon would only travel in interspace material. For N > 0,
𝑖 𝑖−1
( 𝐷−𝑟)+𝑑×𝐼𝑁𝑇( )+𝐷×𝐼𝑁𝑇( )
2 2
𝑎𝑖 = 𝑖 ∈ 1, 2, 3, … , 2𝑁
∆𝑥
(4.39)
For category C8: if N = 0, the photon would only travel in strip material. For N > 0,
𝑖 𝑖−1
( 𝐷+𝑑−𝑟)+𝐷×𝐼𝑁𝑇( )+𝑑×𝐼𝑁𝑇( )
𝑎𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 (4.40)
∆𝑥
𝑖 𝑖−1
( 𝐷+𝑑−𝑟)+𝐷×𝐼𝑁𝑇( )+𝑑×𝐼𝑁𝑇( )
𝑎𝑖 = 2 2
𝑖 ∈ 1, 2, 3, … , 2𝑁 − 1 (4.41)
∆𝑥
For category C10, the photon would only travel in strip material.
The validation of this new method was performed by comparing grid performances determined
using this new method and current methods. Parallel and focused grids were selected from the
scholarly literature, and their performance was also retrieved for comparison. Theoretical
comparisons of radiation transmissions were made only for parallel grids and used mono-energetic
92
radiation. For the focused grids, Monte Carlo simulations were performed by keeping the
The transmissions of primary (Tp) and scatter (Ts) radiation were used as criteria for grid
performance comparisons, indicating the relative primary radiation and scatter radiation reaching
the image receptor with grid to without grid, respectively. Other grid performance factors, as
discussed in Chapter 2, were combination effects of Tp, Ts, and the scatter-to-primary ratio (SPR)
without grid. Using Tp and Ts for grid performance comparisons had the advantage of avoiding
Focused grids were selected from the literature in which the grid geometric details, Tp, and Ts
were readily retrievable. A parallel grid was selected for theoretical comparisons of radiation
4.3.2. Grid details, phantoms, radiation beam quality, and simulation setups
In the simulation, the virtual mammographic grid was constructed using details of the
mammographic grid that was experimentally evaluated in Carton et al. (2009). This is a moving
grid manufactured by Smit Rontgen (Eindhoven, The Netherlands) and contained in a breast
support. The virtual general grid was constructed using the general grid that was experimentally
evaluated in Fetterly and Schueler (2009) and made with lead strips and carbon fibre interspace
materials. For theoretical comparisons, the design of a parallel grid was selected from Dick and
Motz (1978). This grid design was used by Bernstein et al. (1983) for theoretic comparison. The
93
Table 4.3. Details of anti-scatter grids
Carton et al. Fetterly and Schueler Dick and Motz
(2009) grid (2009) grid (1978) grid
Strip thickness d (cm) 0.0020 0.0069 0.0076
Strip material Lead Lead Lead
Interspace distance D (cm) 0.0300 0.0330 0.0174
Interspace material Fibre Fibre Aluminium
Strip height h (cm) 0.1500 0.4950 0.2090
Grid total cover thickness
0.0400 0.0600 0
(cm)
Cover material Fibre Fibre Nil
Focal distance f0 (cm) 65 100
Strip frequency (cm-1) 31 25 40
Grid ratio 5:1 15:1 12:1
The simulations used the same radiation beam qualities as in the experiments in the literature. Rh-
target beams of 35 peak kilovoltage (kVp), 40 kVp, 45 kVp, and 49 kVp with a total 0.27 mm
copper filtration were generated in simulations for the mammographic grid. These beams were
calculated using the model of Boone et al. (1997). They had higher average photon energy than a
common mammographic x-ray beam because they were specially selected for contrast-enhanced
The RQR-8 x-ray beam (International Electrotechnical Commission, 2005) was used in
simulations for the general grid evaluation and was calculated using the model of Poludniowski et
al. (2009b).
When a grid moves during image acquisition, the grid reciprocates about its central line. The
simulation of a grid’s movement was performed in a way in that the grid’s x-coordinate (Figure
94
4.2) changed as a function of time while its y- and z-coordinates remained unchanged. The grid’s
maximum displacement from its central line was set at ±1 cm, the frequency of the grid’s
reciprocation was set at 300 per minute, and the radiation exposure duration was set at 50 ms. The
where Δxgrid is the change of grid x-coordinate, L is the maximum displacement from
grid central line, t is the time at which the photon reaches the grid entry plane, RPM is the
frequency of grid reciprocation, int(A) is the largest integer equal to or smaller than ‘A’,
and Mantissa(A) is the mantissa of ‘A’.
All focused grids were placed at their focal distances from the x-ray focal spot. The parallel grid
was placed at 100 cm from the x-ray focal spot. All grids were placed as close as possible to the
phantoms’ exit-planes.
Water phantoms with a 30 cm × 30 cm cross-sectional area and a range of thicknesses were used
for the general grid. Mammographic phantoms comprising 50% ICRU-44 adipose tissue and 50%
ICRU-44 breast tissue (ICRU, 1989) were used for the mammographic grid. These mammographic
phantoms had a 22 cm × 11.5 cm cross-sectional area and a range of thicknesses (2–8 cm in 2-cm
increments).
An array of detectors (40 cm × 40 cm for the general grid and 24 cm × 30 cm for the
mammographic grid) was placed behind the grid with a 1 cm space to allow the placement of a
moving grid. These detectors had a pixel pitch of 37.5 µm and had no dead space between adjacent
detectors. They were constructed as perfect detectors that convert all of a photons’ energy to signal.
95
4.3.4. Simulation of photon transport
Simulations of photon transport were performed using the dedicated Monte Carlo code system
developed using the details discussed in Chapter 3. The simulations were performed in Matlab Ver.
R2015b (The MathWorks Inc., Natick, Massachusetts) on a computer using an Intel® Core™ i7–
5600U CPU, 2 cores of 2.6 GHz processors and 16 Gigabyte (GB) random access memory.
Each simulation was repeated ten times to obtain the means and standard deviations of Tp, Ts, and
Tt. In each simulation, the number of primary photons generated at the x-ray focal spot increased
as the phantom thickness increased. In the general grid’s evaluation, approximately 100 million
primary photons were generated for the 10-cm thick phantom. The number of primary photons
was increased to approximately 100 billion for the 50-cm thick phantom.
4.3.5. Results
The theoretical transmissions in the parallel grid are shown in Figure 4.3. They were determined
using a mono-energetic radiation beam uniformly distributed across the x-coordinate and are
The radiation mean transmissions determined using this new method has excellent agreement with
the radiation transmissions determined using the methods of Bernstein et al. (1983) and Day and
Dance (1983). Bernstein et al. (1983) found a sign error in one of the equations of Strid (1976).
After correcting this error, we found that the radiation transmission of Strid (1976) is different
from the other methods at some regions. The difference starts at the angle of which the photon
travelled through one grid unit and has a period of one grid unit. Bernstein et al. (1983) suggested
that this difference was caused by the fourth region identified in the Strid (1976) method. However,
96
we found this difference is caused by ignoring the part of the radiation that was in the fourth region
The maximum different transmission from the Strid (1976) method is approximately 14% for the
70-keV photons (Figure 4.3(b)) and more than 1000% for the 30-keV photons (Figure 4.3(a)). As
the absolute difference (less than 1.2%) in the radiation transmission was small, this difference
Figure 4.3. Radiation transmission as a function of photon incident angle. (a) Transmission of
30-keV mono-energetic photons. (b) Transmission of 70-keV mono-energetic photons.
Figure 4.4 shows a magnified region of the radiation transmission of this new method and the Day
and Dance method. The radiation transmission of this new method is a square-wave form with a
minimum value of approximately 0.004, a maximum value of approximately 0.876, and a mean
value of approximately 0.323. The radiation transmission of the Day and Dance method is a
97
Figure 4.4. Transmission of 70-keV mono-energetic beam in a parallel grid in a magnified
region. The entry angle of the beam is 3 degrees. The beam is uniformly distributed across
the grid strips.
The transmissions of non-uniform poly-energetic radiation in the parallel grid are about the same
in this new method and the Day and Dance method (Table 4.4). Tp is approximately 0.585 (SD <=
A parallel grid can be considered symmetrical about any of its strips, assuming the grid has an
infinite size. Whether a parallel grid moves or not, there is always the same relationship between
the strips and the central ray of the primary radiation beam, and there is always grid cut-off in
parallel grids. Despite the parallel grid’s operation (moving or stationary), its radiation
transmissions were the same and therefore agreed with the theoretical transmissions.
Tp Ts
Mean SD Mean SD
The new method (stationary grid) 0.585 0.001 0.082 < 0.001
The new method (moving grid) 0.585 0.001 0.082 < 0.001
Day and Dance method 0.585 < 0.001 0.082 < 0.001
98
4.3.5.2. Transmission of primary radiation in focused grid: Tp
Table 4.5 shows the results of the mammographic grid’s Tp for the 35-kVp beam. Regardless of
the grid’s operation (moving or stationary) and the methods, Tp increased marginally as the
phantom thickness increased. When this grid moved, Tp increased negligibly from 0.841 to 0.843
(SD < 0.001) in the new method. When the grid was stationary, Tp increased negligibly from 0.874
to 0.875 (SD ≤ 0.001) in the new method and from 0.874 to 0.875 (SD < 0.001) in the Day and
Dance method.
Table 4.6 shows the results of the general grid’s Tp. Regardless of the grid’s operation (moving or
stationary) and the methods, Tp increased marginally as the phantom thickness increased. When
this grid moved, Tp increased negligibly from 0.672 to 0.679 (SD ≤ 0.002) in the new method.
When the grid was stationary, Tp increased negligibly from 0.719 to 0.727 (SD ≤ 0.003) in the
new method and from 0.721 to 0.729 (SD ≤ 0.002) in the Day and Dance method.
The difference in Tp between the moving grid and the stationary grid was approximately 4% for
the mammographic grid (Table 4.5) or 7% for the general grid (Table 4.6). This reduction in Tp is
ascribed to grid’s defocusing that occurred during the grid’s movement. This reduction is
significant and is due to grid cut-off. Grid cut-off is undesirable absorption of primary radiation
by grid materials. The difference between the grid cut-offs in the general grid and in the
99
Table 4.5. Transmission of primary radiation (Tp) in the mammographic grid
Tp
4 0.874 (SD < 0.001) 0.841 (SD < 0.001) 0.874 (SD < 0.001)
6 0.875 (SD < 0.001) 0.842 (SD < 0.001) 0.875 (SD < 0.001)
8 0.876 (SD 0.001) 0.843 (SD < 0.001) 0.875 (SD < 0.001)
Results of Tp were obtained with the 35-kVp beam.
Tp
Table 4.7 shows the Ts of the mammographic grid with the 35-kVp beam. Regardless of whether
the grid was moving or stationary, the Ts in this new method is approximately the same as the Ts
100
in the Day and Dance method. As the phantom thickness increased from 2 cm to 8 cm, Ts increased
Table 4.8 shows the Ts of the mammographic grid with the 49-kVp beam. The Ts in this new
method was approximately the same for the moving and stationary grids. As the phantom thickness
increased from 2 cm to 8 cm, the Ts in this new method increased from 0.224 to 0.248 (SD ≤ 0.001)
and Ts in the Day and Dance method increased from 0.219 to 0.233 (SD < 0.001). The difference
in Ts between the new method and the Day and Dance method is significant at two standard
deviations.
Table 4.7. Transmission of scatter radiation (Ts) with 35-kVp beam in the
mammographic grid
Ts
Phantom thickness The new method
Day and Dance Method
(cm) Stationary grid Moving grid
2 0.181 (SD 0.001) 0.180 (SD 0.001) 0.180 (SD 0.001)
4 0.186 (SD 0.001) 0.187 (SD < 0.001) 0.187 (SD < 0.001)
8 0.197 (SD 0.001) 0.197 (SD < 0.001) 0.197 (SD < 0.001)
101
Table 4.8. Transmission of scatter radiation (Ts) with 49-kVp beam in the
mammographic grid.
Ts
Phantom thickness The new method
Day and Dance Method
(cm)
Stationary grid Moving grid
2 0.224 (SD 0.001) 0.224 (SD 0.001) 0.219 (SD < 0.001)
4 0.233 (SD 0.001) 0.233 (SD < 0.001) 0.220 (SD < 0.001)
6 0.241 (SD 0.001) 0.241 (SD 0.001) 0.227 (SD < 0.001)
8 0.248 (SD 0.001) 0.248 (SD < 0.001) 0.233 (SD < 0.001)
Table 4.9 shows the general grid’s Ts. Whether the grid was moving or stationary, Ts was
approximately the same for the new method and the Day and Dance method. As the phantom
thickness increased from 20 cm to 50 cm, Ts increased marginally from 0.063 to 0.066 (SD ≤
0.004). Ts, however, was approximately 0.065 (SD ≤ 0.002) for the 10-cm phantom in both
methods.
Ts
102
4.3.5.4. Images behind a stationary grid and moving grid
Figure 4.5 shows small portions of the 20-cm thick phantom’s images. The images in Figures 4.5(a)
and 4.5(b) were produced using this new method with the grid stationary and moving, respectively.
The image in Figure 4.5(c) was produced using the Day and Dance method with the grid stationary.
Figure 4.5(a) shows the image with grid lines. Figure 4.5(b) and 4.5(c) show the images without
grid lines.
When grids operated when stationary, this new method preserved the grid lines which were
apparent in the image (Figure 4.5). The Day and Dance method, however, eliminated these grid
lines because this method only calculated grid unit mean transmissions, which are not the radiation
transmissions of stationary grids. In previous grid investigations through Monte Carlo simulation
(Malusek et al., 2003, Ullman et al., 2006, Star-Lack et al., 2009, Lazos and Williamson, 2010,
Sun and Star-Lack, 2010, Cunha et al., 2010, Sisniega et al., 2013, Tomal et al., 2013, Chen et al.,
2015, Chen, 2016), the effect of grid lines and the effect of grid cut-off has been neglected.
103
Figure 4.5. Small portions of the 20-cm thick phantom’s images. (a) and (b) Created using the
new method with the grid stationary and moving, respectively. (c) Created using the Day and
Dance method with the grid stationary.
The mammographic grid is a moving grid, so was evaluated while it was moving. Figures 4.6
shows the Tp of this mammographic grid as a function of tube voltage. Despite the phantom
104
thickness and tube voltage, the mammographic grid’s Tp determined using the new method was
less than that determined through the Day and Dance method. At the same tube voltage, the Tp in
both the new method and the Day and Dance method increased marginally (difference less than
0.4%) as the phantom thickness increased from 2 cm to 8 cm. At the same phantom thickness, the
Tp in both the new method and the Day and Dance method also increased marginally (difference
less than 1%) as the tube voltage increased from 35 kVp to 49 kVp. Regardless of the phantom
thickness, as the tube voltage increased from 35 kVp to 49 kVp, the Tp determined through this
new method and through the Day and Dance method increased from approximately 0.841 to 0.851
(SD < 0.001) and 0.874 to 0.883 (SD < 0.001), respectively. The difference in the Tp between the
new method and the Day and Dance method was due to that the grid cut-off effect that was
0.89
Tp
Tp - 2cm (this work)
Tp - 4cm (this work)
Tp - 6cm (this work)
Tp - 8cm (this work)
Tp - 2cm (Day & Dance method)
Tp - 4cm (Day & Dance method)
0.86
Tp - 6cm (Day & Dance method)
Tp - 8cm (Day & Dance method)
0.83
34 39 44 49
Tube voltage (kVp)
105
For both methods, the mammographic grid’s Ts increased as the phantom thickness and/or tube
voltage increased (Figure 4.7). At the same tube voltage, as the phantom thickness increased from
2 cm to 8 cm, the Ts determined increased approximately 11% in the new method and
approximately 10% in the Day and Dance method. At the same phantom thickness, as the tube
voltage increased from 35 kVp to 49 kVp, the Ts increased approximately 26% in the new method
and approximately 20% in the Day and Dance method. Regardless of the phantom thickness, as
the tube voltage increased from 35 kVp to 49 kVp, the Ts determined through the new method and
the Day and Dance method increased from approximately 0.180 to 0.248 (SD ≤ 0.001) and 0.180
0.26
Ts Ts - 2cm (this work)
0.17
34 39 44 49
Tube voltage (kVp)
The Day and Dance method underestimated the mammographic grid’s Ts for the higher tube kVp
(40, 45, and 49 kVp). At the 35 kVp tube voltage, the Ts in the new method was approximately
the same as the Ts in the Day and Dance method. From 40 kVp to 49 kVp, the Ts in the new
106
method was significantly different from the Ts in the Day and Dance method at two standard
deviations. This significant difference was attributed to two factors: (1) the Day and Dance method
extrapolated focused grids’ strip angles and (2) the Day and Dance method assumed that scatter
4.3.6.1. Tp comparison
In general, the mammographic grid’s Tp increased as the tube voltage increased in both the
simulation and in Carton et al. (2009) (Figures 4.8 and 4.9). Regardless of the method, the
mammographic grid’s Tp obtained from the simulation increased marginally as the phantom
thickness and/or tube voltage increased (Figure 4.8). In Carton et al. (2009), the Tp increased
significantly as the tube voltage increased or the Tp varied marginally as the phantom thickness
The general grid’s Tp obtained through the simulation increased marginally (less than 1%) as the
phantom thickness increased from 10 cm to 30 cm (Figure 4.10). The general grid’s Tp was
approximately constant in Fetterly and Schueler (2009) (Figure 4.10). In the simulation, the Tp
increased as the phantom thickness increased because the beam-hardening effect increased as the
When these grids were operated when moving, the Tp determined through the new method was a
better approximation of Carton et al. (2009) and Fetterly and Schueler (2009) than the Tp
determined through the Day and Dance method (Figures 4.9 and 4.10). This was because the effect
of grid cut-off due to grid movement was accounted for in the new method. When the general grid
107
was operated when stationary, the Tp was approximately the same for the new method and the
0.89
Tp
Tp - 2cm (this work)
0.86
Tp Tp - 2cm (this work)
108
Tp Tp - Fetterly & Schueler, 2009
0.60
10 20 30 40 50
The difference in the mammographic grid’s Tp between the new method and Carton et al. (2009)
decreased from approximately 4% to 1% as the tube voltage increased from 35 kVp to 49 kVp
(Figure 4.9). The difference in the mammographic grid’s Tp between the Day and Dance method
and Carton et al. (2009) decreased from approximately 8% to 4% as the tube voltage increased
from 35 kVp to 49 kVp. The difference in the general grid’s Tp between the new method with the
grid moving and Fetterly and Schueler (2009) was approximately 5.3% (SD = 0.004) (Figure 4.10).
The difference in Tp between the new method with the grid stationary and Fetterly and Schueler
4.3.6.1. Ts comparison
The mammographic grid’s Ts increased in both the simulation and Carton et al. (2009) as the tube
voltage and/or phantom thickness increased (Figures 4.11 and 4.12). The general grid’s Ts
remained approximately the same in the simulation and Fetterly and Schueler (2009) as the
phantom thickness increased (Figure 4.13). This trend of the mammographic grid’s Ts increasing
with increasing tube voltages agrees with Carton et al. (2009) and Salvagnini et al. (2012).
109
0.26
Ts Ts - 2cm (this work)
Ts - 4cm (this work)
0.23 Ts - 6cm (this work)
Ts - 8cm (this work)
Ts - 2cm (Day & Dance method)
0.20
Ts - 4cm (Day & Dance method)
Ts - 6cm (Day & Dance method)
0.17 Ts - 8cm (Day & Dance method)
34 39 44 49
Tube voltage (kVp)
0.00
10 20 30 40 50
Phantom thickness (cm)
110
The difference in the mammographic grid’s Ts between the simulation and the literature decreased
as the tube voltage and/or phantom thickness increased (Figure 4.12). The mean difference in the
mammographic grid’s Ts between this new method and Carton et al. (2009) decreased from
approximately 21% to 13% (SD ≤ 0.066) as the tube voltage and phantom thickness increased
from 35 kVp to 49 kVp and 2 cm to 8 cm, respectively. The mean difference in the mammographic
grid’s Ts between the Day and Dance method and Carton et al. (2009) decreased from
approximately 19% to 10% (SD ≤ 0.085) as the tube voltage and phantom thickness increased
from 35 kVp to 49 kVp and 2 cm to 8 cm, respectively. In Carton et al. (2009), the standard
deviation of the Ts at 40 kVp was greater than 0.03. At two standard deviations, the difference in
the mammographic grid’s Ts between the simulation and Carton et al. (2009) was not significant.
Regardless of the methods or the grid’s operation (moving or stationary), the difference in the
general grid’s Ts between the simulation and Fetterly and Schueler (2009) was approximately 16%
(Figure 4.13).
The difference in Tp or Ts between the simulation and the literature may be ascribed to several
causes, such as grids’ manufacture defects, and/or difficulty in the experimental setup. The exact
causes, however, cannot be determined through either the simulation or the experiment.
Summary
This chapter presented the development and validation of a new method to determine radiation
transmission in grids. This new method overcame the limitations of current methods as set out in
the scholarly literature. These limitations included: (1) assuming scatter radiation is uniformly
distributed in the grid entry surface; (2) extrapolating radiation transmission from grid-unit mean
111
transmission, which does not preserve the details of grid lines resulting from stationary grids; and
The new method was developed using photon path lengths in the grid’s strip and interspace
materials. These lengths were determined using a geometrical constraint, which is a constant x-
coordinate for parallel grids, or a constant x-coordinate to z-coordinate ratio for focused grids. The
new method used these lengths and these materials’ linear attenuation coefficients to determine
the radiation transmission in the grid. This and the other methods all ignored the scatter radiation
Validation of this new method was performed by comparing grid performance determined through
the simulation using the new method and contemporary methods, and that retrieved from the
literature. Most emphasis was made on the differences between the new method and the Day and
Dance method, which is the most appropriate contemporary method for determining radiation
The results of the parallel grid showed excellent agreement between the mean transmission in the
new method and the transmissions of other methods in the scholarly literature, except Strid’s (1976)
method. Strid’s method designated four mutually exclusive regions to determine the radiation
transmission in grids. Bernstein et al. (1983) suggested that Strid’s method wrongly designated the
fourth region. However, we found that Strid’s method did neglect transmission of radiation located
in the fourth region and only passed partially through the grid strip.
The comparisons of radiation transmissions in the focused grids were made only between the
literature, the new method, and the Day and Dance method. The results of the focused grids showed
112
that Tp was affected by grid operation. When the grids were operated when stationary, the Tp
determined using the new method was approximately the same as that obtained through the Day
and Dance method and was higher than that in the literature. When the grids were operated when
moving, the Tp determined using the new method was approximately 4% for the mammographic
grid and 7% for the general grid, which is lower than that obtained through the Day and Dance
method and higher than that in the literature. The difference in Tp between this new method and
the Day and Dance method was due to the grid-cut off effect of the grids’ movement. The Day and
Dance method did not account for the grid-cut off effect on Tp.
Ts was not affected by the grid’s operation. The general grid’s Ts determined using the new
method was approximately the same as that obtained using the Day and Dance method and was
higher than that in the literature. The mammographic grid’s Ts determined using the new method
was higher than that obtained using the Day and Dance method and was higher than that in the
literature. The Day and Dance method, however, underestimated the mammographic grid’s Ts.
The discrepancy of the mammographic grid’s Ts between the new method and the Day and Dance
When the grids were operated when stationary, the Day and Dance method did not preserve the
grid lines because their method only extrapolated radiation transmission from the grid-unit mean
transmission, which is not the radiation transmission of stationary grids. The new method
preserved these grid lines which were shown as white strips in the image.
The discrepancy in Tp and Ts between the simulation and the literature may be ascribed to several
causes, such as grids’ manufacturing defects, the difference between the simulation setup and the
113
experimental setup, and/or difficulty in the experimental setup. The exact causes, however, cannot
Chapter 2 to Chapter 4 presented the details for using our Monte Carlo simulation code system to
evaluate grid designs. Chapter 5 will show the validation of our Monte Carlo simulation code
system.
114
Chapter 5: Validation of the new Monte Carlo code
system for grid evaluation
This chapter presents the final part of the first phase of the project. It validates the Monte Carlo
simulation code system that was developed using the details presented in Chapter 3. The validation
is done through grid performance comparisons between simulation results and experiment results.
This validation was presented at the Engineering & Physical Science in Medicine 2016 conference.
Electrotechnical Commission, 2013) international standards, which are recommended for the
experimental evaluation of grids. The setups were made under three conditions: (1) a narrow beam
condition for primary radiation measurement, (2) a broad beam condition with a primary beam
blocker for scatter radiation measurement, and (3) a broad beam condition for total radiation
measurement. The details of these setups were discussed in Chapter 2. The measurements of the
amount of radiation obtained in these setups were used to calculate four quantities: The
transmissions of primary (Tp), scatter (Ts), total (Tt) radiation, and the scatter-to-primary ratio
(SPR) without grid. Tp, Ts and Tt were calculated using Equations 2.1 to 2.3 (see Chapter 2). SPR
𝑆𝑐𝑎𝑡𝑡𝑒𝑟_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑔𝑟𝑖𝑑
𝑆𝑃𝑅 = (5.1)
𝑃𝑟𝑖𝑚𝑎𝑟𝑦_𝑅𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑔𝑟𝑖𝑑
115
In addition, a range of the primary beam’s field of views (FOVs) was simulated to investigate the
effect of the FOV on Tp. A range of primary beam blocker sizes was also simulated to investigate
the effect of blocker size on Ts. In the experiment, only one FOV of primary beam and one primary
In the simulation, logic labels were assigned to these primary photons which, without interaction,
would travel within predefined primary beam FOVs and/or within predefined primary beam
blocker areas. A logic label occupies one byte of random access memory (RAM). These labels
used only relatively negligible amounts of RAM during the simulation. By using photon labels,
the simulation setup was simplified to the broad beam condition, which was used for all
measurements.
The advantage of using labels for photons is that simulation time can be reduced by one-third
because one setup is sufficient to obtain measurements of total and scatter radiation, which
otherwise can only be obtained experimentally in two separate setups. Seven primary beam FOVs
and seven primary beam blocker sizes were investigated in the simulation.
In the experiment, the amount of radiation was measured using a DRX Evolution Plus 3543 Image
Rochester, NY, USA). The detailed information on the construction of this image receptor is
proprietary; as such, it could not be obtained. In the simulation, a Gd2O2S image receptor
(phosphor height 600 μm, phosphor area 50 µm × 50 µm, silicon structure thickness 40 µm) was
constructed for the approximation of the DRX Evolution Plus 3543 Image Receptor. In each
radiation measurement setup, the experiment and simulation were repeated three times to obtain
116
three sets of measurements. These repeated sets of measurements were used to calculate means
To compare the experiment and the simulation, tube voltages were selected depending on phantom
thickness. These voltages were selected because x-ray imaging generally uses tube voltage
depending on the thickness of the imaging anatomy: the thicker the anatomy, the higher the tube
voltage. In the experiment and the simulation, voltages increased from 60 to 100 peak kilovoltage
increments.
X-ray beam spectrums used in the simulation were calculated using the model of Poludniowski et
al. (2009b). The quality of radiation beams used in the experiment and the simulation was
measured (Table 5.1). The half-value layers of radiation beams used in the simulation were
approximately 0.1 mm lower than those used in the experiment, but this difference would have
had negligible effects on the simulation and experimental results. The beams’ mean energy used
in the simulation was measured, but not that used in the experiment.
117
5.1.4. Grid details
Grid details are listed in Table 5.2. The grid was received from Philips Medical Systems, Germany
and was made of lead strips, carbon fibre interspace, and aluminium covers. The paint materials
on the grid entry and exit surfaces have a thickness of approximately 1.44 mm. As recommended
by the grid’s manufacturer, in the simulation, these materials were weighted by increasing the strip
5.1.5. Phantoms
Polymethyl methacrylate (PMMA) was used as the phantom material. Each piece of PMMA had
a 30 cm × 30 cm cross-section area and 1 cm thickness. The phantoms’ thicknesses were set from
Inc., Natick, Massachusetts) on a computer using an Intel® Core™ i7–5600U CPU, 2 cores of 2.6
118
GHz processors and 16 Gigabyte RAM. Radiation transmissions in the grid materials were
The number of primary photons generated at the x-ray focal spot increased as the phantom
thickness increased. Approximately 100 million primary photons were generated for the 10-cm
thick phantom. The number of primary photons was increased to approximately 10 billion for the
5.2. Results
5.2.1. Results obtained through simulation
As the radius of the primary beam’s FOV increased from 0.15 cm to 2 cm in the simulation, Tp
determined with the same kVp was approximately the same (Figure 5.1). As the kVp and phantom
thickness increases from 60 kVp and 10 cm to 100 kVp and 30 cm, respectively, Tp increased
from approximately 0.636 to 0.716. The statistic variation of Tp decreased as the radius of the
primary beam’s FOV increased because the larger the primary beam size resulted in more photons,
119
Tp
60 kVp & 10 cm
0.74
70 kVp & 15 cm
0.72
0.64
0.62
0.60
0 0.5 1 1.5 2
Radius of primary beam's FOV at detector (cm)
In the simulation, Ts determined with the same kVp decreased as the radius of the primary beam’s
FOV increased from 0.15 cm to 2 cm (Figure 5.2). Ts decreased from approximately 0.048 to
0.033 for the 60 kVp and 10 cm phantom, 0.062 to 0.045 for the 70 kVp and 15 cm phantom, 0.071
to 0.060 for the 80 kVp and 20 cm phantom, 0.087 to 0.074 for the 90 kVp and 25 cm phantom,
and 0.097 to 0.090 for the 100 kVp and 30 cm phantom. The statistical variation of Ts decreased
as the radius of the primary beam’s FOV increased due to the same reason as that for the statistic
variation of Tp.
120
0.14
Ts 60 kVp & 10 cm
90 kVp & 25 cm
0.08
100 kVp & 30 cm
0.06
0.04
0.02
0.00
0 0.5 1 1.5 2
Radius of primary beam's FOV at detector (cm)
Figure 5.2. Ts as a function of primary beam size. Ts decreases as the radius of primary
beam’s FOV increases. The error bar indicates two standard deviations.
In the simulation, Tt determined with the same kVp was approximately the same as the radius of
the primary beam’s FOV increased from 0.15 cm to 2 cm (Figure 5.3). As the kVp and phantom
thickness increase from 60 kVp and 10 cm to 100 kVp and 30 cm, respectively, Tt decreased from
approximately 0.222 to 0.134. The statistic variation of Tt decreased as the radius of the primary
beam increased due to the same reason as that for the statistical variation of Tp.
121
Tt
60 kVp & 10 cm
0.24
70 kVp & 15 cm
0.22
80 kVp & 20 cm
0.20
90 kVp & 25 cm
0.16
0.14
0.12
0.10
0 0.5 1 1.5 2
Radius of primary beam's FOV at detector (cm)
Figure 5.3. Tt as a function of tube voltage and phantom thickness. Tt remains approximately
unchanged as the radius of primary beam’s FOV increases. The error bar indicates two
standard deviations.
In the simulation, as the radius of the primary beam increased from 0.15 cm to 2 cm (Figure 5.2),
SPR determined with the same kVp decreased from approximately 2.320 to 2.175 for the 60 kVp
and 10 cm phantom, 4.556 to 3.964 for the 70 kVp and 15 cm phantom, 7.129 to 6.297 for the 80
kVp and 20 cm phantom, 11.196 to 9.259 for the 90 kVp and 25 cm phantom, and 18.925 to 13.050
for the 100 kVp and 30 cm phantom. The statistical variation of SPR decreased as the radius of
the primary beam increased due to the same reason as for the statistical variation of Tp.
122
SPR
60 kVp & 10 cm
20
18 70 kVp & 15 cm
16
80 kVp & 20 cm
14
12 90 kVp & 25 cm
10
100 kVp & 30 cm
8
0
0 0.5 1 1.5 2
Radius of primary beam's FOV at detector (cm)
Figure 5.4. SPR as a function of tube voltage and phantom thickness. SPR decreases as the
radius of primary beam’s FOV increases. The error bar indicates two standard deviations.
As the tube voltage and phantom thickness increased from 60 kVp and 10 cm to 100 kVp and 30
cm, respectively, Tp determined in the experiment increased from approximately 0.623 to 0.718
(Figure 5.5). Ts increased from approximately 0.047 to 0.088 (Figure 5.6), Tt decreased from
approximately 0.219 to 0.135 (Figure 5.7), and SPR increased from approximately 2.452 to 13.065
(Figure 5.8).
123
0.74 Tp
0.72
0.70
0.68
0.66
0.64
0.62
0.60
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.5. Tp determined in experiment as a function of tube voltage and phantom thickness.
Tp increases as the tube voltage and phantom thickness increase. The error bar indicates two
standard deviations.
0.10
Ts
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.6. Ts determined in experiment as a function of tube voltage and phantom thickness.
Ts increases as the tube voltage and phantom thickness increase. The error bar indicates two
standard deviations.
124
Tt
0.24
0.22
0.20
0.18
0.16
0.14
0.12
0.10
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.6. Tt determined in experiment as a function of tube voltage and phantom thickness.
Tt decreases as the tube voltage and phantom thickness increase. The error bar indicates two
standard deviations.
14.00 SPR
12.00
10.00
8.00
6.00
4.00
2.00
0.00
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.8. SPR determined in experiment as a function of tube voltage and phantom thickness.
SPR increases as the tube voltage and phantom thickness increase. The error bar indicates
two standard deviations.
125
5.3. Discussion
As the tube voltage and phantom thickness increase, the trend of Tp increasing in the simulation
agrees with that in the experiment (Figure 5.9). This increasing trend is due to the increased
average photon energy because the higher the photon energy, the lower the radiation attenuation
in the grid materials. As the tube voltage and phantom increased from 60 kVp and 10 cm to 100
kVp and 30 cm, respectively, the difference in Tp between the simulation and the experiment
Tp experiment Simulation
0.74
0.72
0.70
0.68
0.66
0.64
0.62
0.60
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Tube voltage and phantom thickness
Figure 5.9. Tp as a function of tube voltage and phantom thickness. Tp increases as the tube
voltage and phantom thickness increase. The error bar indicates two standard deviations.
As the tube voltage and phantom thickness increase, the trend of Ts increasing in the simulation
agrees with that in the experiment (Figure 5.10). This increasing trend is also due to the increased
126
average photon energy because the higher the photon energy, the lower the radiation attenuation
in the grid materials. As the tube voltage and phantom increased from 60 kVp and 10 cm to 102
kVp and 30 cm, respectively, the difference in Ts between the simulation and the experiment
decreased from approximately 20.0% to 1.1%. The difference in Ts between the simulation and
the experiment with the 60 kVp and 10 cm phantom and the 70 kVp and 15 cm phantom is
significant at two standard deviations. This difference may be caused by difficulty in the
experiment setup as shown in section 5.2.1 that the primary beam’s blocker size affects Ts (Figure
5.2).
0.08
0.07
0.06
0.05
0.04
0.03
0.02
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.10. Ts as a function of tube voltage and phantom thickness. Ts increases as the tube
voltage and phantom thickness increase. The error bar indicates two standard deviations.
Tt obtained through the simulation is approximately the same as that determined in the experiment
(Figure 5.11). As the tube voltage and phantom thickness increase, the trend of Tt decreasing in
127
the simulation agrees with that in the experiment. The difference in Tt between the simulation and
Tt experiment Simulation
0.24
0.22
0.20
0.18
0.16
0.14
0.12
0.10
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.11. Tt as a function of tube voltage and phantom thickness. Tt decreases as the tube
voltage and phantom thickness increase. The error bar indicates two standard deviations.
SPR is approximately the same in the simulation and the experiment (Figure 5.12). As the tube
voltage and phantom thickness increase from 60 kVp and 10 cm to 100 kVp and 30 cm, the SPR
increases from approximately 2.452 to 13.065 in the experiment and 2.278 to 12.592 in the
simulation. The difference in SPR between the simulation and the experiment is approximately,
128
SPR experiment Simulation
14.00
12.00
10.00
8.00
6.00
4.00
2.00
0.00
60 kVp & 10 cm 70 kVp & 15 cm 80 kVp & 20 cm 90 kVp & 25 cm 100 kVp & 30 cm
Figure 5.12. SPR as a function of tube voltage and phantom thickness. SPR increases as the
tube voltage and phantom thickness increase. The error bar indicates two standard deviations
In general, the results obtained through the simulation are approximately the same as those
determined in the experiment. Some discrepancy in Ts between the simulation and the experiment
In summary, our Monte Carlo simulation code system is appropriate for analysing and/or
evaluating the designs of grids which can be, or have been, manufactured with negligible defects.
Summary
This chapter presented the validation of our Monte Carlo simulation code system for grid
evaluation. This validation was made by comparing grid performance determined through the
simulation and the experiment. The grid performance was evaluated using the beam-block method,
which was reviewed in Chapter 2. The simulation was performed using approximately the same
conditions as that in the experiment. In addition, the simulation also investigated additional
129
conditions: a range of primary beam sizes for the measurement of primary radiation and a range
Comparing grid performances between the simulation and the experiment showed that the Tp, Ts,
Tt, and SPR determined in our Monte Carlo simulation code system were approximately the same
as those determined in the experiment. The investigation of a range of primary beam sizes in the
simulation showed that the Tp was not affected by the primary beam’s FOV. The investigation of
a range of primary beam blocker size showed that the Ts decreased as the radius of the primary
beam blocker increased. This is because the more the primary beam was blocked, the less forward
scatter radiation that arose from the phantom’s matter underneath the blocker.
In this chapter, our Monte Carlo simulation code system was proven that it is appropriate for
analysing and/or evaluating the designs of grids that can be, or have been, manufactured with
negligible defects. In the next chapter, a new method to design new grids was developed to
130
Chapter 6: Criteria for designing new grids
This chapter describes the second phase of the project. It gives an overview of current grid designs
and then discusses factors and issues related to grid design. The final part of this chapter evaluates
Two grid design theories are reported in the scholarly literature, each of them focusing on different
radiation transmissions of grids. One theory (Strid, 1976) directs grid design towards the maximum
reduction of scatter radiation by optimising the grid strip thickness. In the previous chapter, the
transmission of scatter radiation (Ts) was found that it depends on tube voltage. This means an
‘optimal’ strip thickness at a certain tube voltage may no longer be ‘optimal’ when the tube voltage
changes.
The other theory (Bonenkamp and Hondius Boldingh, 1959b) presents a mechanism to maximise
the transmission of primary radiation (Tp) for a given grid lead content, which is measured as mass
per unit area (commonly in units g/cm2). In this theory, Tp is increased by reducing the strip
thickness and keeping the grid-unit length unchanged (a grid-unit length is the sum of strip
thickness and an adjacent interspace distance). As the grid lead content is constant, reducing the
strip thickness increases the strip height. Consequently, the grid ratio is increased. In this theory,
for a given lead content, the best grid would be the one constructed with the largest grid ratio.
131
These theories focus on optimising Ts or maximising Tp, but not both. In the following sections
of this chapter, a new grid design method was investigated to optimise Ts and maximise Tp
simultaneously.
Grid design factors are these parameters that affect Tp and/or Ts. There are two types of factors.
One type relates to grid construction, including strip thickness, strip height, strip material,
interspace material, interspace distance, grid ratio, grid cover material, and grid cover thickness.
The other type relates to grid application, including x-ray beam quality, scatter radiation
distribution, and grid application limits. The relationship between these factors is illustrated in
Figure 6.1: the factor at each arrow’s end influences the factor at the arrow’s head.
Figure 6.1. Relationships between grid design factors for the transmission of scatter radiation
and primary radiation. The factor at the arrow’s head is influenced by the factor at this arrow’s
end.
132
Grid ratio is the determining factor for grid application limits, which are the distances at which Tp
40% for general application grids and 20% for mammographic grids (International
As the grid ratio determines a grid’s application limits, in the following analysis of the relationship
illustrated in Figure 6.1, it is assumed that the grid ratio is constant. Equivalently, the grid
The material and distance of the grid interspace are the essential factors for maximising Tp. These
factors’ effects on Ts should be negligible because allowing the primary radiation to travel through
the interspace region would also permit the scatter radiation to travel through this region.
The height, material, and thickness of the grid strip are crucial factors for optimising Ts. The
thickness affects Tp because primary radiation travelling inside strips is completely absorbed.
Since primary radiation does not travel through grid strips, the combination effect of strip material
and height on Tp can be ignored. Strip height, however, determines the interspace height which
In the previous chapter, Tp and Ts were found that they increase as the tube voltage increases.
Tube voltage is a determining factor for x-ray beam quality, which together with the imaging
anatomy determines the distribution of scatter radiation. The beam quality directly affects Tp. The
133
6.3. Issues in grid design
Issues in grid design relate to the determinations of Tp and Ts, which are the basic grid
performance factors. These factors can be determined through experiments with prototype grids;
however, manufacturing prototype grids is not only expensive but also takes a long time. In many
situations, such as designing new grids or simulating x-ray imaging systems with grids,
When the directions and locations of the photons on the grid entry surface are known, the
transmission of these photons in the grid’s materials can be determined. In a 3-dimension Cartesian
coordinate system, a photon (P) with energy ɛ enters a grid from its entry surface at an arbitrary
point (x, y, z) with a direction (Δx, Δy, Δz). This photon, without interaction with the grid, would
exit this grid and reach a point S(x0, y0, z0) in the image receptor’s area. The integration of the
energy of all photons without interacting with the grid materials and reaching point S can be
expressed as Equation 6.1. If the grid were not there, the integration of the energy of all these
photons reaching the image receptor area could be expressed as Equation 6.2.
Neglecting the effects of these photons’ directions and the radiation detector’s materials on energy
deposition in the image receptor, Equation 6.1 determines the total energy at the point S with the
grid and Equation 6.2 determines the total energy at the point S without the grid. These effects can
be accounted for in these equations by including the angular factors of these directions and the
Primary photons’ coordinates and directions at the grid’s entry surface can be analytically
determined. However, scatter photons’ coordinates and directions at the grid’s entry surface cannot
be analytically determined. This is because scatter photons have randomly, non-uniformly arisen
134
from the imaging anatomy. Without knowing the scatter photons’ coordinates and directions at the
grid’s entry surface, Ts cannot be determined using Equations 6.1 and 6.2.
𝐸(𝑥0 , 𝑦0 , 𝑧0 ) =
∫ ∫ ∫ ∫ ⟨𝑃(𝜀, 𝑥, 𝑦, 𝑧, ∆𝑥, ∆𝑦, ∆𝑧)|𝑃(∆𝑥, ∆𝑦, ∆𝑧) ≡ 𝑆(𝑥0 , 𝑦0 , 𝑧0 )⟩ 𝑒 −𝑢𝑠 (𝜀)∗𝑙𝑠 −𝑢𝑖 (𝜀)∗𝑙𝑖 𝑑𝜀𝑑𝑧𝑑𝑦𝑑𝑥
(6.1)
(6.2)
where < A | B > is a function that returns ‘A’ if ‘B’ is true, otherwise returns zero,
P(ɛ,x, y, z, Δx, Δy, Δz) returns the photon P’s energy ɛ, P(Δx, Δy, Δz) ≡ S(x0, y0, z0) returns
true if and only if aΔx = x0, aΔy = y0, and aΔz = z0 with ‘a’ being a positive real number,
µs(E) and µi(E) are the linear attenuation coefficients of the strip material and interspace
material for photons of energy E, respectively, and ls and li are the distances that the photons
will travel in the strip material and interspace material without interaction, respectively.
The Monte Carlo simulation of x-ray imaging can determine an approximation of the distribution
of scatter photons at the grid entry surface. The accuracy of this approximation, however, can be
increased by increasing the number of total photons in the simulation (Landau and Binder, 2009).
In this project, the Tp and Ts of grid designs were determined through Monte Carlo simulations.
Grids should be designed to substantially remove scatter radiation reaching the image receptor
(minimising Ts), and to also maximally retain primary radiation reaching the image receptor
(maximising Tp). As primary radiation is generated at a finite x-ray focal spot, the primary
135
radiation’s beam reaching the grid entry surface is a divergent beam. Since scatter radiation arises
randomly and non-uniformly from any point in the imaging anatomy, the scatter radiation reaching
Focused grids permit maximal transmission of primary radiation as the strips of these grids are
aligned to the divergent primary beam. They may be linear or cross-hatched. A linear focused grid
has its strips parallel and aligned to focus on a common line. Contemporary grids are almost always
focused. In the following discussion, focused grids with a given grid ratio are assumed.
Tp depends on four factors: the strip thickness, the interspace material, the interspace distance, and
the strip height. Strid (1976) has found that Ts is optimised by an optimal strip thickness. In
Chapter 2, a criterion was proposed for determining strip optimal thickness. Strip thickness,
The interspace region is best as a vacuum. Alternatively, it can be air-filled for easy grid
construction.
Since the grid ratio is fixed, a change in the interspace distance must be paired with a proportional
change in the strip height. As the strip thickness is constant, when the interspace distance is
increased, the Tp will increase. Tp then can be increased by concurrently increasing the interspace
distance and the strip height without changing the grid ratio. Alternatively, Tp depends on the strip
height.
Ts depends on four factors: scatter radiation distribution, strip material, strip height, and strip
thickness. The scatter radiation distribution, which depends on both the imaging anatomy and the
x-ray beam quality, varies from one x-ray examination to another. The scatter radiation distribution
136
is set for a standard imaging condition (International Electrotechnical Commission, 2013, 2001,
The strip material should be high atomic number material, such as lead or tungsten. However, lead
is highly poisonous; if it is inhaled or swallowed, it damages the nervous system and causes brain
disorders. Therefore, tungsten strips are more desirable for all types of grids. Molybdenum strips
increased, Ts should not depend on the strip height. Ts, then, primarily depends on the strip
thickness which, as discussed above, can be optimised. The strip optimal thickness can be
After the interspace distance and strip height are proportionally increased, the volume of the
imaging anatomy (the green region in Figure 6.2) is increased. The scatter radiation arising in this
volume may travel directly towards the interspace underneath this volume and through the grid
without any interaction with the grid’s materials. If the grid ratio increases, this volume will
decrease. Therefore, this volume can be kept approximately unchanged or decreased by increasing
137
Figure 6.2. Relationships between strip thickness (d), interspace (D), strip height (h), and front
and back cover thicknesses (x1 and x2, respectively).
Based on the above, there are three essential criteria for designing new grids with a given grid ratio:
Summary
The prime goal of the second phase of this research was to develop a new method to design new
grids to overcome the reduction of primary radiation in grid materials. In this chapter, an overview
of two current contemporary grid design theories was presented. One of these theories, the Strid’s
grid design theory, emphasises on optimising grid strip thickness to maximise grid performance.
The other theory, the Bonenkamp and Hondius Boldingh’s grid design theory, focuses on
138
In this chapter, the factors and issues in grid design were also analysed. Then new criteria were
determined for designing new grids: (1) using air-interspace materials; (2) increasing interspace
distance and the strip height, proportionally; (3) optimising strip thickness. These provide guidance
for designing new air-interspaced grids whose Tp and Ts will only depend on their strip height and
strip optimal thickness, respectively. Such new grids could overcome the problem of the reduction
In the next chapter, such new grid designs were evaluated through Monte Carlo simulation to
determine a solution for the reduction of scatter radiation without increasing patients’ radiation
139
Chapter 7: Monte Carlo evaluation of new grid designs
and performance results
This chapter describes the last phase of the project, using all the achievements made in the previous
phases to evaluate a solution for avoiding the reduction of primary radiation reaching the image
receptor. A series of new grid designs was constructed and evaluated through Monte Carlo
simulations. This chapter has three sections: the first provides the details of material and methods
used for the evaluation; the second presents the evaluation results; and the third compares these
New mammographic grid designs were made with reference to scatter-to-primary ratio (SPR), tube
voltages, and grid ratios in current mammography, where SPR is approximately 0.2–1.3 (Barnes
and Brezovich, 1978, Boone and Cooper III, 2000, Carton et al., 2009, Zhou et al., 2016), tube
voltages may vary from approximately 24 peak kilovoltage (kVp) to 49 kVp (Carton et al., 2009,
Cunha et al., 2010), and mammographic grids have a lead strip around 20 μm thick and grid ratios
A series of mammographic grid designs was made with a range of grid ratios and a range of strip
thicknesses. The combination of these ratios and thicknesses gave a total of 781 grids. The grids
were designed with tungsten strips and air interspace, and the strip height was fixed at 2 cm taller
141
than contemporary mammographic grids’ strip heights. This height would increase the penumbra
effect. Therefore, a justification was made to reduce the potential penumbra effect by increasing
the grid focal distance so that the image’s lack of sharpness ascribed to the penumbra effect was
New general grid designs were made with reference to SPR, tube voltages, and grid ratios in
current general x-ray examinations, where SPR might vary from less than 0.1 (such as for a finger
x-ray) to 26 (such as for a 50-cm thick abdominal x-ray) (Fetterly and Schueler, 2007), tube
voltages of general x-ray beams have a broad range (such as 50 kVp for wrist x-ray or 120 kVp
for a chest x-ray), general grid ratios may vary from 5:1 to 21:1 (Fetterly and Schueler, 2009), and
grid lead strip thickness may vary from 30 µm to more than 100 μm (Strid, 1976, Fetterly and
Schueler, 2009).
A series of general grid designs were made with a range of grid ratios and strip thicknesses. The
combination of these ratios and thicknesses gave a total of 1800 grids. These grids were designed
142
with tungsten strips and air interspace, and the grid height was fixed at 5 cm. This height was
justified for current contemporary Bucky systems, which have a space in between the exit-surface
of the imaging anatomy and the entry surface of the image receptor. This space is intentionally
designed using the air-gap technique to improve the efficiency of scatter radiation reduction (Bell
et al., 2003). It is approximately 10 cm and allows an adequate space for these new grids to be
successfully assembled in Bucky systems. The details of these grids are listed in Table 7.2.
Two virtual breast phantoms of semi-circular cross-section were used in the simulation. The cross-
section had an 8-cm radius. These phantoms were composed of, in sandwich structure, 50%
glandular tissue and 50% adipose tissue using ICRU (1989) standards (Table 7.3).
In the literature, a 5-cm thick breast phantom is often used for grid performance evaluation, a
image diagnostic quality of thin breasts was found not to benefit from mammographic grids, but
143
the image diagnostic quality of thick (approximately 5 cm or more) breasts was found to benefit
Total thickness 2 cm 8 cm
In this project, two phantoms were selected to represent the thinnest (2 cm) and thickest (8 cm)
breasts. If a grid could improve the image diagnostic quality of both the thinnest breast and the
thickest breast, this grid should benefit the image diagnostic quality of all breasts.
The justification for using 50% glandular tissue, 50% adipose tissue, and semi-circular cross-
section phantoms was that this phantom’s composition was the closest approximation of breasts
of various thicknesses from 2 cm to 8 cm (Cunha et al., 2010). SPRs from different phantom
material have been found to be different (Dance et al., 2000, Huda et al., 2003, Cunha et al., 2010,
Cunha et al., 2012, Dance et al., 2012, Tomal et al., 2013, Sisniega et al., 2013). Ts changes as
SPR changes (Carton et al., 2009, Salvagnini et al., 2012, Zhou et al., 2016). Using approximate
breast phantoms would minimise any potential effects on Ts due to the difference between the
144
7.1.2.2. Virtual water phantoms for general grid evaluation
Two virtual water phantoms were used to evaluate new general grid designs. One phantom was 10
cm thick, which represents a thin anatomy such as an ankle. The other phantom was 30 cm thick,
which represents a thick anatomy such as an abdomen. These phantoms had a cross-section of 30
cm × 30 cm.
The justification for using water phantoms was that they approximated body soft tissue
composition. A 20-cm thick water phantom is recommended for general grid evaluation (Strid,
1976, Fetterly and Schueler, 2007, International Electrotechnical Commission, 2013, International
cm and 30 cm thicknesses were selected in this study because if the images of both the thin and
thick anatomies could benefit from a grid, this grid should be of benefit to the image quality
thicker than 50 cm for a very large person. However, this project did not use a phantom thicker
than 30 cm because the 30 cm is the thickness of most thick imaging anatomies: the thicknesses
of most lateral spines is approximately 30 cm and the thicknesses of extremities, head and body
Another reason for not using a phantom thicker than 30 cm was to save simulation time. To achieve
a similar level of statistic error, the time needed for the simulation of a 40-cm thick anatomy can
be 20 times more than that needed for a 30-cm thick anatomy. In this project, one simulation of a
145
7.1.3. X-ray beam quality
Ts increases as tube voltage increases (Carton et al., 2009, Zhou et al., 2016). In mammography,
tube voltages are selected depending on breast thickness. The highest possible tube voltages of x-
ray beams for breast thicknesses were selected for evaluating the new mammographic grid designs.
These x-ray beams were calculated using the model of Boone et al. (1997). The beams’ details
28 2 17.7 0.311
36 8 19.5 0.404
In general radiography, tube voltages are generally selected depending on the imaging anatomy
thickness: the greater the thickness, the higher the tube voltage. In the simulation, tube voltages
were increased as the phantom thickness increased. X-ray beams were calculated using TechnicVR
146
Table 7.5. Details of x-ray beam
kVp Phantom thickness (cm) Mean energy (keV) Half-value layer Al (mm)
60 10 cm 38.5 2.8
image, and the image is read using an imaging plate reader. The reader uses a specific laser to
stimulate the phosphors in the plate to emit light which then is detected by a photomultiplier tube.
Digital mammography/radiography uses various types of image receptors. These image receptors
form part of either a directly detectable imaging system or an indirectly detectable imaging system.
In indirectly detectable imaging systems, radiation interacts principally with a phosphor screen.
Immediately after an interaction with radiation, the phosphors in the screen emit visible light which
is then detected by a two-dimensional array of photodiodes (Yaffe and Rowlands, 1997). Image
receptors used in indirect detectable imaging systems are a Gd2O2S:Tb (Gadolinium oxysulphide)
or CsI (caesium iodide scintillator) phosphor screen coupled with photodiodes (Zhao et al., 2004).
polycrystalline layers of HgI2, PbO, or CdZnTe, is used as the principal detecting element. The
photoconductor absorbs x-ray photons and converts their energy directly to electronic charges
(Kasap et al., 2011). The a-Se detectors are suitable for mammographic x-ray beams because they
have a high detection quantum efficiency for these beams (Kasap et al., 2011).
147
An amorphous selenium image receptor was constructed in the simulation for the mammographic
grid evaluation, with an 85 µm pitch and a 200 μm thick selenium layer. It had 2816 × 3584 pixels,
was 240 mm × 300 mm, and was like a flat-panel detector used in the Siemens MAMMOMAT
A CsI/Tl (550 µm thickness) image receptor was constructed in the simulation for the general grid
evaluation. It was 43 cm × 43 cm, 3000 × 3000 pixels, and was like the Pixium 4600 CsI/Tl flat-
panel image receptor (Trixell, Moirans, France), whose performance was evaluated in Geijer et al.
(2001).
were not used. Primary or scatter photons were distinguished by their marker, primary or scatter,
respectively. The simulation of photon transportation was performed in Matlab Ver. R2015b (The
MathWorks Inc., Natick, Massachusetts) on a computer using an Intel® Core™ i7–5600U CPU,
2 cores of 2.6 GHz processors and 16 Gigabyte RAM. Radiation transmissions in the grid materials
The simulation of each grid evaluation for each phantom was repeated 10 times to obtain 10 sets
All simulation results were fitted using an interpolant from Matlab Ver. R2015b (The MathWorks
Inc., Natick, Massachusetts) with a biharmonic (v4) method. This interpolant uses the “griddata”
function in the Matlab for the biharmonic method and fits an interpolating surface that passes
148
through every data point. The fitting statistics were summarised using two parameters: the sum of
squared error (SSE) and the coefficient of determination (R-squared). SSE is also known as the
sum of the squared difference between each measurement, and the measurements’ mean and is a
measure of variation within the measurements (Durbin and Watson, 1951). SSE can be determined
using Equation 7.1. R-squared is the square of the correlation coefficient, which measures the
degree of a linear relationship between variables. R-squared can be used to analyse the degree of
connection between variables (Sharma et al., 1981) and can be determined using Equation 7.2
Where ∑𝑛𝑖=1 𝑎𝑖 is the sum of 𝑎𝑖 for indices 𝑖 = 1 to 𝑖 = n, 𝑥𝑖 is the ith sample of the 𝑥
variable, and 𝑥 is the mean of the samples.
2
𝑛×∑𝑛𝑖=1(𝑥𝑖 ×𝑦𝑖 )−∑𝑛𝑖=1 𝑥𝑖 ×∑𝑛𝑖=1 𝑦𝑖
R_squared = [ 2 2
] (7.2)
√[∑𝑛𝑖=1 𝑥𝑖 2 −(∑𝑛𝑖=1 𝑥𝑖 ) ]×[∑𝑛𝑖=1 𝑦𝑖 2 −(∑𝑛𝑖=1 𝑦𝑖 ) ]
Where ∑𝑛𝑖=1 𝑎𝑖 is the sum of 𝑎𝑖 for indices 𝑖 = 1 to 𝑖 = n; 𝑥𝑖 is the ith sample of the 𝑥
variable; 𝑦𝑖 is the ith sample of the 𝑦 variable; the ith sample of the 𝑥 variable is paired with
the ith sample of the 𝑦 variable; n is the total number of paired samples.
7.2. Results
7.2.1. Mammographic grid results
The fitting statistics of the mammographic grid designs’ Tp are, for both phantoms, the SSE less
149
Tp decreased as the grid ratio or/and the strip thickness increased (Figures 7.1 and 7.2). As the
grid ratio and the strip thickness increased from 5:1 to 15:1 and 2 μm to 50 μm, respectively, Tp
decreased in the 2-cm phantom from 0.993 to 0.958 (SD < 0.001), and in the 8-cm phantom from
Figure 7.1. Tp for the 2-cm phantom as a function of the strip thickness and grid ratio. Tp was
fitted with a surface plane.
150
Figure 7.2. Tp for the 8-cm phantom as a function of the strip thickness and grid ratio. Tp was
fitted with a surface plane.
The fitting statistics of the new mammographic grid designs’ Ts were calculated. The SSE was
0.004 for the 10-cm phantom and 0.001 for the 30-cm phantom. The R-squared was greater than
Ts decreased as the grid ratio or/and strip thickness increased (Figures 7.3 and 7.4). As the grid
ratio and strip thickness increased from 5:1 to 15:1 and 2 μm to 50 μm, respectively, Ts decreased
in the 2-cm phantom from 0.352 to 0.052 (SD < 0.001) and in the 8-cm phantom from 0.521 to
151
Figure 7.3. Ts for the 2-cm phantom as a function of the strip thickness and grid ratio. Ts was
fitted with a surface plane.
Figure 7.4. Ts for the 8-cm phantom as a function of the strip thickness and grid ratio. Ts was
fitted with a surface plane.
152
7.2.1.3. Mammographic grids: KSNR
KSNR is the quantum signal-to-noise ratio (SNR) improvement factor (KSNR). The fitting statistics
of the mammographic grid designs’ KSNR were calculated. The SSE was less than 0.001 for the
10-cm phantom and 0.002 for the 30-cm phantom. The R-squared was greater than 0.999 for both
phantoms.
For a given grid ratio, KSNR first increased then decreased as the strip thickness increased (Figures
7.5 and 7.6). For a thin strip (e.g. 2 µm), KSNR increased as the grid ratio increased from 5:1 to
15:1. For a thick strip (e.g. 40 µm), KSNR first increased then decreased as the grid ratio increased.
As the grid ratio increased from 5:1 to 15:1 at the strip optimal thicknesses, KSNR increased in the
2-cm phantom from 1.257 to 1.277 (SD < 0.001) and in the 8-cm phantom from 1.992 to 2.135
153
Figure 7.5. KSNR for the 2-cm phantom as a function of the strip thickness and grid ratio. KSNR
was fitted with a surface plane.
Figure 7.6. KSNR for the 8-cm phantom as a function of the strip thickness and grid ratio. KSNR
was fitted with a surface plane.
154
7.2.1.4. Strip optimal thicknesses of mammographic grids
In Section 2.2, a criterion was discussed for the determination of strip optimal thickness for a given
strip height and grid ratio. This criterion was used to determine the mammographic grid designs’
With the same phantom thickness, the strip optimal thickness decreased as the grid ratio increased
(Figures 7.7 and 7.8). With the same grid ratio, the strip optimal thickness increased as the phantom
thickness increased. As the phantom thickness and tube voltage increased together, the increase of
the optimal thickness was ascribed to the combination effect of the phantom thickness and the tube
voltage. It is anticipated that the strip optimal thickness is affected by the tube voltage, the phantom
As the grid ratio increased from 5:1 to 15:1 at these strip optimal thicknesses, Tp changed less than
1%; Ts decreased in the 2-cm phantom from 0.168 to 0.058 and in the 8-cm phantom from 0.198
to 0.069; and KSNR increased in the 2-cm phantom from 1.257 to 1.277 and in the 8-cm phantom
155
Figure 7.7. Tp and Ts for the 2-cm phantom as functions of the strip thickness and grid ratio.
Both Tp and Ts were fitted with surface planes.
Figure 7.8. Tp and Ts for the 8-cm phantom as functions of the strip thickness and grid ratio.
Both Tp and Ts were fitted with surface planes.
156
7.2.2. General grid results
The fitting statistics of these general grid designs’ Tp were calculated. The SSE was 0.007 for the
10-cm phantom and 0.005 for the 30-cm phantom. The R-squared was greater than 0.999 for both
phantoms.
Tp decreased as the grid ratio and/or the strip thickness increased (Figures 7.9 and 7.10). Tp
decreased from 0.994 to 0.894 (SD = 0.002) as the grid ratio and strip thickness increased from
5:1 to 40:1 and 2 μm to 140 μm, respectively. The difference in Tp between the two phantoms was
Figure 7.9. Tp for the 10-cm phantom as a function of the strip thickness and grid ratio. Tp was
fitted with a surface plane.
157
Figure 7.10. Tp for the 30-cm phantom as a function of the strip thickness and grid ratio. Tp
was fitted with a surface plane.
The fitting statistics of these general grid designs’ Ts were calculated. The SSE was 0.006 for the
10-cm phantom and 0.001 for the 30-cm phantom. The R-squared was greater than 0.999 for both
phantoms.
Ts decreased as the grid ratio and/or the strip thickness increased. As the grid ratio and strip
thickness increased from 5:1 to 40:1 and 2 μm to 140 μm, respectively, Ts decreased in the 10-cm
phantom from 0.835 to 0.023 (SD < 0.001) and in the 30-cm phantom from 0.844 to 0.025 (SD <
0.001).
158
Figure 7.11. Ts for the 10-cm phantom as a function of the strip thickness and grid ratio. Ts
was fitted with a surface plane.
Figure 7.12. Ts for the 30-cm phantom as a function of the strip thickness and grid ratio. Ts
was fitted with a surface plane.
159
7.2.2.3. General grids: KSNR
The fitting statistics of these general grid designs’ KSNR were calculated. The SSE was 0.128 for
the 10-cm phantom and 1.277 for the 30-cm phantom. The R-squared was greater than 0.999 for
both phantoms.
For a given grid ratio, KSNR first increased then decreased as the strip thickness increased (Figures
7.13 and 7.14). For a thin strip (for example 2 µm), KSNR increased as the grid ratio increased from
5:1 to 40:1. For a thick strip (for example 130 µm), KSNR first increased then decreased as the grid
ratio increased. As the grid ratio increased from 5:1 to 40:1 at the strip optimal thicknesses, KSNR
increased in the 10-cm phantoms from 2.786 to 3.303 (SD = 0.01) and in the 30-cm phantom from
Figure 7.13. KSNR for the 10-cm phantom as a function of the strip thickness and grid ratio.
KSNR was fitted with a surface plane.
160
Figure 7.14. KSNR for the 30-cm phantom as a function of the strip thickness and grid ratio.
KSNR was fitted with a surface plane.
The strip optimal thicknesses of these general grid designs were determined using the criterion
With the same phantom thickness, the general grid designs’ strip optimal thickness decreased as
the grid ratio increased (Figures 7.15 and 7.16). With the same grid ratio, the strip optimal
thickness increased as the phantom thickness increased. The trend of the strip optimal thickness
was the same for the mammographic grid designs and the general grid designs. As the tube voltages
for the general grid designs’ evaluation were also selected depending on phantom thickness, the
general grid designs’ strip optimal thickness should also be affected by the tube voltage, the
161
As the grid ratio increased from 5:1 to 40:1 at these strip optimal thicknesses (Figures 7.15 and
7.16), Tp decreased in the 10-cm phantom from 0.983 to 0.978 and in the 30-cm phantom from
0.974 to 0.971; Ts decreased in the 10-cm phantom from 0.206 to 0.029 and in the 30-cm phantom
from 0.311 to 0.029; KSNR increased in the 10-cm phantom from 2.786 to 3.303 and in the 30-cm
Figure 7.15. Tp and Ts for the 10-cm phantom as functions of the strip thickness and grid ratio.
Tp and Ts were fitted with surface planes.
162
Figure 7.16. Tp and Ts for the 30-cm phantom as functions of the strip thickness and grid ratio.
Tp and Ts were fitted with surface planes.
7.3. Discussion
7.3.1. Mammographic grids
The Tp and Ts of the new mammographic grid designs with strip optimal thicknesses were
compared with the literature. Despite the grid ratios and/or phantom thicknesses, the Tp of the new
mammographic grid designs were approximately 23% higher than those reported by Carton et al.
(2009), or 33% higher than those reported by Salvagnini et al. (2012) (Figures 7.17 and 7.18).
163
Tp 1
Tp - 2cm phantom
0.9
Tp - 2cm phantom & 35kVp (Carton
et al. 2009)
Tp - 2cm phantom & Inspiration
0.8 system (Salvagnini et al., 2012)
0.7
4 5 6 7 8 9 10 11 12 13 14 15
Grid ratio
Tp 1
Tp - 8cm phantom
0.9
Tp - 8cm phantom & 35kVp (Carton
et al. 2009)
Tp - 8cm phantom & Inspiration
0.8 system (Salvagnini et al., 2012)
0.7
4 5 6 7 8 9 10 11 12 13 14 15
Grid ratio
164
The Ts of the new mammographic grid designs depended on the grid ratio and/or phantom
thickness. The dependence of Ts on phantom thickness agrees with the findings of Carton et al.
(2009). Contemporary mammographic grids are commonly built with a grid ratio of 5:1 (Carton
et al., 2009, Cunha et al., 2010, Salvagnini et al., 2012). The Ts of the grid ratio 5:1 for the new
mammographic grid designs was higher than those of Carton et al. (2009) and Salvagnini et al.
(2012) (Figures 7.19 and 7.20). For the 2-cm phantom, the new mammographic grid design’s Ts
was approximately 20% higher than Carton et al. (2009) and approximately 14% higher than
Salvagnini et al. (2012) (Figure 7.19). For the 8-cm phantom, the new mammographic grid
design’s Ts was approximately 18% higher than Carton et al. (2009) and 27% higher than
Ts
Ts - 2cm phantom
0.05
4 5 6 7 8 9 10 11 12 13 14 15
Grid ratio
165
Ts
Ts - 8cm phantom
0.05
4 5 6 7 8 9 10 11 12 13 14 15
Grid ratio
The difference in Ts between the grid ratio of 5:1 for the new mammographic grid design and
those in Carton et al. (2009) and Salvagnini et al. (2012) may be due to several causes: the
difference in the x-ray beam quality, the difficulty in the experimental setups, the difference in
these grids’ strip height, or any combination of these. The effect of the strip height on Ts can be
counteracted by increasing the grid ratio: a grid ratio of 6:1 for the new mammographic grid design
has approximately the same Ts as those of Carton et al. (2009) and Salvagnini et al. (2012).
Despite the phantom thicknesses, the Ts of the grid ratio of 15:1 of the new mammographic grid
design was approximately 57% lower than Carton et al. (2009). For the 2-cm phantom, this grid
design’s Ts was approximately 60% lower than Salvagnini et al. (2012). For the 8-cm phantom,
this grid design’s Ts was approximately 46% lower than Salvagnini et al. (2012).
166
7.3.1.2. New mammographic grid designs vs the literature: KSNR
The KSNR of the new mammographic grid designs with strip optimal thicknesses were compared
with the literature. The KSNR of the new mammographic grid designs were greater than 1.26 for
the 2-cm phantom (Figure 7.21) or greater than 1.99 (1.99 to 2.14) for the 8-cm phantom (Figure
7.22). The KSNR of those grids in Carton et al. (2009) and Salvagnini et al. (2012) were 0.82 and
0.90, respectively, for the 2-cm phantom (Figure 7.21) and 1.005 and 1.06, respectively, for the 8-
The KSNR of the new mammographic grid designs were 40 to 43% higher than (Carton et al., 2009)),
and 52 to 55% higher than Salvagnini et al. (2012) for the 2-cm phantom (Figure 7.21). The KSNR
of these new mammographic grid designs were 89 to 102% higher than (Carton et al., 2009)) and
98 to 113% higher than Salvagnini et al. (2012) for the 8-cm phantom (Figure 7.22).
In the literature, mammographic grids were found to improve only thick breasts’ image diagnostic
quality, when breast thickness was larger than approximately 5 cm (Bernhardt et al., 2006). The
KSNR of the new mammographic grid designs were greater than 1.26, regardless of the phantom
thickness and/or grid ratio. New mammographic grid designs, however, will improve image
diagnostic quality for all breasts, whether they are thin or thick.
167
1.4 KSNR
1.3
1.2
2cm phantom
1.1
2cm phantom & 35kVp
1 (Carton et al. 2009)
0.7
4 5 6 7 8 9 10 11 12 13 14 15
Grid ratio
KSNR
2.1
1.9
8cm phantom
1.7
8cm phantom & 35kVp
1.5 (Carton et al. 2009)
0.9
4 5 6 7 8 9 10 11 12 13 14 15
Grid ratio
168
7.3.2. General grids
The Tp and Ts of the new general grid designs with strip optimal thicknesses were compared with
the literature. Despite the phantom thickness and/or grid ratio, the Tp of the new general grid
designs were approximately the same (less than 1% difference) (Figures 7.23 and 7.24). The Tp of
Kim et al. (2007) and Fetterly and Schueler (2007, 2009) decreased linearly as the grid ratio
increased (Figures 7.23 and 7.24). The Tp of the new general grid designs were 0.974 to 0.983,
which was 31 to 38% higher than Kim et al. (2007) and 29 to 43% higher than Fetterly and Schueler
(2007, 2009).
Tp - 10cm phantom
0.95
0.85
0.75
0.65
5 15 25 35
Grid ratio
169
Tp - 30cm phantom
0.95
0.85
0.75
0.65
5 15 25 35
Grid ratio
There was a trend for the Ts of the new general grid designs to decrease as the grid ratio increased.
As the grid ratio increased from 5:1 to 41:1, the Ts of these new grid designs decreased from
approximately 0.21 to 0.03 in the 10-cm phantom (Figure 7.25) and approximately 0.31 to 0.03 in
the 30-cm phantom (Figure 7.26). This trend agrees with Fetterly and Schueler (2007), Fetterly
Regardless of the phantom thickness, for the same grid ratio, the new grid designs’ Ts were on
average 56% (23 to 64%) less than Kim et al. (2007), 32% (24 to 40%) less than Fetterly and
Schueler (2007) and 17% (7 to 27%) less than Fetterly and Schueler (2009) (Figures 7.25 and 7.26).
The best Ts of the new grid designs was on average 92% (91 to 94%) less than Kim et al. (2007),
80% (73 to 88%) less than Fetterly and Schueler (2007) and 61% (50 to 72%) less than Fetterly
170
Ts - 10cm phantom
Ts - 20cm phantom (Kim et al., 2007)
Ts - 20cm phantom (Fetterly et al., 2007, 2009)
Ts
0.5
0.4
0.3
0.2
0.1
0
5 15 25 35
Grid ratio
Ts - 30cm phantom
Ts - 20cm phantom (Kim et al., 2007)
Ts - 20cm phantom (Fetterly et al., 2007, 2009)
Ts
0.5
0.4
0.3
0.2
0.1
0
5 15 25 35
Grid ratio
171
7.3.2.2. New general grid designs vs the literature: KSNR
The KSNR of the new general grid designs with strip optimal thicknesses were compared with the
literature. The KSNR of the new general grid designs increased as the grid ratio increased (Figures
7.27 and 7.28). This trend of increasing KSNR agrees with Fetterly and Schueler (2007), Fetterly
In the 10-cm phantom, KSNR of the new grids were 2.79 to 3.30 (Figure 7.27). In the 30-cm
phantom, KSNR of the new grids were 5.67 to 10.07) (Figure 7.28). Regardless of the phantom
thicknesses, the new general grid designs’ KSNR were 35 to 635% higher than Fetterly and Schueler
(2007, 2009). The new general grid designs’ KSNR were several times (2.1 to 11.4) more than Kim
et al. (2007).
10cm phantom
50cm phantom (Fetterly et al., 2007, 2009)
20cm phantom (Fetterly et al., 2007, 2009)
20cm phantom (Kim et al., 2007)
KSNR
3.5
2.5
1.5
0.5
5 10 15 20 25 30 35 40
Grid ratio
172
30cm phantom
50cm phantom (Fetterly et al., 2007, 2009)
20cm phantom (Fetterly et al., 2007, 2009)
20cm phantom (Kim et al., 2007)
10 KSNR
8
6
4
2
0
5 10 15 20 25 30 35 40
Grid ratio
Summary
The discoveries made in the first and second phases of this research were applied in this chapter
to determine a solution for the reduction of primary radiation in grid materials. In this chapter,
many new grid designs were created and evaluated through the Monte Carlo simulation.
The simulation results showed that it is possible to have the Tp of new grid designs approximately
equal to the Tp of a perfect grid (Tp = 1) without compromising the new grid designs’ Ts, which
depends on the grid ratio – the higher the grid ratio, the smaller the Ts. The application of the new
grid designs to scatter radiation reduction will reduce negligible amounts of primary radiation: Tp
is more than 0.99 for the mammographic grid designs and 0.974 to 0.983 for the general grid
designs. When compared to contemporary grids in the literature, new grids using these new grid
designs will reduce the stochastic effects of ionising radiation, due to the compensation for the
173
primary radiation reduction in the grid materials, by approximately 39% for general radiography
With the same grid ratios, new grids using these new grid designs will have a much smaller Ts
than those of contemporary grids in the literature, except for the grid ratio of 5:1 for the new
mammographic grid design. Furthermore, scatter radiation reaching the image receptor can be
further reduced by using such new grids with a high grid ratio, without increasing the stochastic
effects.
With all new grid designs, the KSNR increases as the imaging anatomy thickness increases and/or
the grid ratio increases. For the thinnest breast (2 cm thickness), the KSNR of the new
mammographic grid designs is 1.26 to 1.28. For the 10-cm thick anatomy, the KSNR of the new
general grid designs is 2.79 to 3.30, which is greater than the KSNR of those grids in the literature
for all phantom thicknesses (10 to 50 cm). For the same grid ratio and the same anatomical
thickness, the KSNR of the new general grid designs is more than three times greater than those in
the literature.
The KSNR of all the new grid designs implies that new grids using these new grid designs will have
the benefit of improved image diagnostic quality, regardless of anatomical thicknesses, either in
found to only improve image diagnostic quality for breasts thicker than approximately 5 cm.
The KSNR of these new grid designs showed that for a given SPR, grid performance is a
combination effect of Tp and Ts. A grid with a high KSNR may not have the lowest Ts; however, a
grid with the lowest Ts among grids with the same Tp will always have the highest KSNR.
174
Building grids using these new designs can provide an effective ionising radiation reduction
This means that the lowest radiation exposure can be achieved using such new grids to produce x-
175
Chapter 8: Conclusions
In this research, new designs of anti-scatter grids were investigated. Grids are commonly used in
medical x-ray imaging systems to reduce scatter radiation reaching the image receptor. The strips
of grids reduce primary radiation transmitting through them and leading to increase patient
exposure to ionising radiation. New grid design criteria were revealed to maximise the
transmission of primary radiation (Tp) for a given grid ratio to avoid increasing patient exposure
to ionising radiation. A novel criterion for the determination of optimal thickness of grid strips was
investigated to maximise grid performance. A new method for radiation transmission in grid
materials was also developed for grid evaluation in Monte Carlo simulation. This method
determines the transmission probability of each photon instead of the grid-unit mean transmission
determined by the contemporary methods. These new grid designs were evaluated using a Monte
Carlo simulation code system which was developed and validated in this research. The results
showed that grids developed with these new designs would reduce scatter radiation reaching the
image receptor without increasing patient exposure to ionising radiation. Replacing current grids
by grids using these new designs, more scatter radiation would be stopped from reaching the image
receptor. In addition, the patient would be exposed to less ionising radiation, approximately
The investigation of new grid designs was carried out in three distinct phases. In the first phase
(Chapters 2 to 5), grid evaluation methodology through Monte Carlo simulation was reviewed and
investigated. The development of a new grid design method was completed in the second phase
177
(Chapter 6). In the final phase (Chapter 7), the achievements made in Chapters 2 to 6 were applied
to determine the transmissions of primary (Tp), scatter (Ts), and total radiation (Tt) of those new
designs. Signal to noise ratio (SNR) improvement factor (KSNR) was derived from these
transmissions. The comparisons of Tp, Ts, Tt, and KSNR between these new grid designs and the
current grids investigated in the literature were also presented in the last phase.
In Chapter 1, an overview of the background and the opportunity for this research on the anti-
scatter grid technique were provided. This technique is used in most medical x-ray imaging
systems which incorporate the use of ionising radiation to provide clinicians with a diagnosis for
their patients. This research focused on grids used in mammography and general radiography.
Mammography and general radiography produce images of which diagnostic quality is formed by
the differentiation attenuation arising when primary radiation interacts with the matter of tissues
or organs.
An interaction between primary radiation and matter can scatter the primary radiation and produce
scatter radiation. Generally, through the interaction, primary radiation ionises the atoms in the
tissues and organs. Ionisation in living tissues and organs causes detrimental effects, including
deterministic effects and stochastic effects. The deterministic effects occur when a patient’s
radiation exposure exceeds the threshold dose. The stochastic effects have no threshold dose. The
probability of the occurrence of the stochastic effects increases as a patient’s cumulative radiation
dose increases. The linear-no-threshold (LNT) theory predicts that the probability of the
occurrence of the stochastic effects is proportional to the cumulative ionising radiation dose
received by a patient.
178
In mammography and general radiography, an image is formed by the intensities of both the
attenuated primary radiation beam and the scatter radiation reaching the image receptor. Images
are commonly stored and viewed in digital forms by many rows and columns of pixels. In a digital
x-ray image, the difference between the pixel values of adjacent regions is the image contrast,
which is reduced by scatter radiation reaching the image receptor. As clinicians rely on image
contrast to make confident and accurate diagnoses from images, scatter radiation in an image,
especially the images of thick imaging anatomies (such as an adult abdomen x-ray image), is a
major problem affecting clinicians’ confidence in making accurate diagnoses. This problem still
exists in mammography and general radiography, as well as other x-ray imaging technologies.
A common solution for removing scatter radiation reaching the image receptor is to use an anti-
scatter grid technique. This technique has been used for nearly a century in most x-ray imaging
systems. Other scatter radiation reduction techniques (the slot technique and air-gap technique) are
not commonly used because of their limitations. Some ‘gridless’ techniques (software solutions)
are currently under development where researchers attempt to study scatter radiation and subtract
it from an image to improve the image’s diagnostic quality. Some applications of these software
solutions have been used in bedside radiography. While the improvement in image diagnostic
quality by these software solutions is not as good as that from the anti-scatter grid technique, these
solutions are useful due to difficulties in using grids when performing bedside radiography.
Grids reduce a great portion of scatter radiation reaching the image receptor. Using grids, however,
still results a relatively great ratio of scatter radiation to primary radiation in images of thick
anatomies. When an x-ray image is undertaken for a 20-cm thick anatomy (such as an adult
abdomen radiography) without using a grid, the scatter-to-primary ratio (SPR) in the image could
179
be as much as 5.9. If a contemporary grid is used, the SPR could still be approximately 1.3. This
means that even if a grid is used, the intensities of the scatter radiation in this image are still about
Grids reduce primary radiation reaching the image receptor. The reduction of primary radiation
increases the image Poisson mottle, which makes it difficult to make a diagnosis from the image.
Radiation exposures are increased to avoid increasing the image Poisson mottle. The increased
radiation exposure, however, increases the stochastic effects of ionising radiation delivered to the
patients. When grids are used, the increase in the stochastic effects due to the increased radiation
exposure to compensate the reduction of primary radiation in the grid materials is approximately
Patient radiation protection guides the practice of using ionising radiation in medical diagnostic or
treatment applications to prevent the occurrence of the deterministic effects and minimise
stochastic effects. Using grids that increase stochastic effects of ionising radiation raises concerns
for radiation protection in relation to the as low as reasonably achievable (ALARA) principle. This
principle requires x-ray examinations to be performed using the lowest possible radiation dose
delivered to the patients and without compromising the image diagnostic quality. A challenge in
scatter radiation reduction for mammography and general radiography is to substantially remove
scatter radiation reaching the image receptor, without increasing radiation exposure to the patients
This problem, that using grids to reduce scatter radiation reaching the image receptor increases the
stochastic effects of ionising radiation delivered to the patients, gave rise to the research in this
180
thesis. This project provided a solution to reduce patient exposure to ionising radiation without
compromising image diagnostic quality by developing and evaluating new grid designs.
In Chapter 2, grid evaluation criteria and methods were reviewed. These criteria and methods were
used to evaluate grid performance in the experiment performed in Chapter 5 and through the Monte
Carlo simulations performed in Chapters 4, 5, and 7. In Chapter 2,the basics grid criteria: Tp, Ts,
and Tt, and KSNR were discussed. Then the setups for the radiation measurements to determine
these criteria using a beam block method was examined. This method determines primary, scatter,
and total radiation using three measurement conditions: narrow beam condition for primary
radiation, broad beam condition with a primary beam blocker for scatter radiation, and broad beam
condition for total radiation. Furthermore, a novel criterion, the first differential of the KSNR, was
analysed to determine the optimal strip thickness for a given grid ratio and strip height. Analysis
showed that the KSNR increases as the strip thickness increases from zero and decreases as the strip
thickness increases further. This novel criterion assumed that there is a maximum KSNR for strip
thickness from zero to infinity. This criterion was used in Chapter 7 to determine the optimal strip
In Chapter 3, details for developing our Monte Carlo simulation code system for grid evaluation
were reviewed and discussed. After the principles for photon transportation in matter were
reviewed, a simulation procedure for the simulation of x-ray imaging was developed. In this
procedure, photon transportation was tracked from the x-ray focal spot to the image receptor. The
details of the events in this procedure were analysed for: (1) changes in a photon’s location,
direction, and/or energy; (2) random number generation and sampling methods; and (3)
interactions between photons and matter; and (4) radiation dosimetry and virtual phantom
181
construction. Our Monte Carlo simulation code system was developed following this simulation
procedure.
The changes in a photon’s location, direction, and/or energy were determined using linear
attenuation coefficients and random numbers. Linear attenuation coefficients were retrieved from
XCOM data base maintained by the National Institute of Standards and Technology, USA.
Uniform random numbers were generated using the Mersenne Twister algorithm, which has a
period of length 219937 − 1. Non-uniform random numbers were sampled using either reject-accept
Three types of interactions, Rayleigh scattering, Compton scattering, and photoelectric absorption,
were implemented in our Monte Carlo simulation code system. The scatter photon directions
depended on the atomic form factor for Rayleigh scattering and on the incoherent scattering
function for Compton scattering. If the interference effect of compounds on the scatter photon
direction was accounted for, the scatter photon’s direction was also dependent on the interference
function for Rayleigh scattering. The Doppler effect on the scatter photon’s energy broadening
was ignored as this effect was found negligible; however, this effect could be implemented using
a probability distribution of the energy broadening. In all the interactions, the kinetic energy of
emitted electrons was locally deposited in the matter. The tracking of electrons was ignored.
Virtual phantoms were constructed using either the polynomial boundaries of materials or the
In Chapter 4, a new method was developed and validated to overcome the limitations set out in the
analytical methods in the literature for determining photon transportation in grid materials. These
analytical methods are used in Monte Carlo simulation of photon transportation to efficiently
182
determine radiation transmission in grid materials. These limitations included: (1) assuming scatter
radiation is uniformly distributed in the grid entry surface; (2) extrapolating radiation transmission
from grid-unit mean transmission, which does not preserve the details of grid lines resulting from
stationary grids; and (3) extrapolating radiation transmission for focused grids. This new method
used photons’ path lengths in strip and interspace materials and these materials’ linear attenuation
The validation of this new method was made by comparing grid performance determined through
simulation using this new method and contemporary methods, and that retrieved from the literature.
These comparisons showed that this new method and the Day and Dance method, which is the
most appropriate analytical method for determining radiation transmission in grid materials,
determined approximately the same Tp in both parallel grids and focused grids. However, they
also showed that the Day and Dance method underestimated the mammographic grid’s Ts. The
discrepancy in the mammographic grid’s Ts between this new method and the Day and Dance
method increased as the tube voltage increased. Grids can be operated while moving or stationary.
In the simulations, when the grids were operated when stationary, the Day and Dance method did
not preserve the grid lines because the method only extrapolated radiation transmission from the
grid-unit mean transmission which is not the radiation transmission of stationary grids. But our
new method preserved these grid lines which were shown as white strips in the image. The
comparisons also showed discrepancies in Tp and Ts between the simulation and the literature.
These differences may be ascribed to several causes, such as grids’ manufacturing defects, the
difference between the simulation setup and the experiment setup, and/or difficulty in the
experiment setup.
183
From Chapter 2 to Chapter 4, the details of using a Monte Carlo simulation code system to evaluate
grid designs were studied. These details were used to develop our Monte Carlo code system. This
system was validated through the comparisons of grid performance determined in a simulation and
In Chapter 5, an experiment and a simulation were performed to determine SPR, Tp, Ts, and Tt of
one anti-scatter grid. Results showed that the Tp, Ts, Tt, and SPR determined in the simulation
were approximately the same as those determined in the experiment. In addition, through
investigating a range of primary beam’s field-of-view (FOV) and primary beam’s blocker size, it
showed that primary beam’s FOV only negligibly affects Tp. Ts, however, is affected by the
primary beam’s blocker size due to the scatter radiation that would have been arisen from the
phantom’s material underneath the blocker. These results demonstrated that our Monte Carlo
The primary goal set in the second phase of this research was accomplished in Chapter 6. A new
grid design method was developed to overcome the reduction of primary radiation in grid materials.
As found in Chapter 1, current grids have not been designed to maximise Tp and optimise Ts. In
Chapter 6, grid design factors and issues were analysed and then new criteria for designing new
grids were determined. These new criteria are: (1) using air-interspace materials; (2) increasing
interspace distance and the strip height, proportionally; and (3) optimising strip thickness. These
criteria provided guidance for designing new air-interspaced grids whose Tp and Ts will only
depend on their strip height and strip optimal thickness, respectively. Such new grids could
184
The evaluation results of the grid new designs were presented in Chapter 7. The achievements
made in the first and second phases were used to evaluate the designs of many new grids through
Monte Carlo simulation. The results from this evaluation were presented in terms of Tp, Ts, and
KSNR. These results affirmed a solution for reducing scatter radiation reaching the image receptor
When a grid is used during imaging, the grid’s Tp is an indication of a patient’s exposure to
ionising radiation: the less the Tp, the greater the patient’s exposure to ionising radiation. The Tp
of new grid designs were approximately equal to the Tp of a perfect grid (Tp = 1) without
compromising these new grid designs’ Ts. Generally, Ts depends on the grid ratio, the higher the
grid ratio, the smaller the Ts. The application of these new grid designs to scatter radiation
reduction would reduce negligible amounts of primary radiation as Tp is more than 0.99 for the
mammographic grid designs and more than 0.97 (0.974 to 0.983) for the general grid designs.
Using new grids made from these new designs to replace contemporary grids would reduce the
stochastic effects of ionising radiation due to the compensation for the primary radiation reduction
in the grid materials. The reduction of these stochastic effects would be approximately 39% for
KSNR is an indication of the improvement in image quality. A KSNR equal to 1 means that the grid
would not improve the image quality. A KSNR less than 1 means that the grid would reduce the
image quality; the smaller the KSNR, the greater the reduction. A KSNR greater than 1 means that
the grid would improve the image quality; the higher the KSNR, the greater the image quality
improvement.
185
The KSNR of new mammographic grid designs increased from 1.257 to 1.277 in the 2-cm phantom
and 1.992 to 2.135 in the 8-cm phantom, as grid ratio increased from 5:1 to 15:1. The KSNR of new
mammographic designs was more than 40 to 43% higher than that in the literature for the 2-cm
phantom and more than 89 to 102% higher than that in the literature for the 8-cm phantom. With
the same grid ratios, new grids made from these new mammographic designs would improve the
image quality for all breast thicknesses. In contrast, current mammographic grids were found they
only improve the image quality for breast thicknesses approximately greater than 5 cm.
The KSNR of new general grid designs increased from 2.786 to 3.303 in the 10-cm phantom and
5.670 to 10.073 in the 30-cm phantom, as the grid ratio increased from 5:1 to 40:1. Regardless the
phantom thicknesses, with the same grid ratio, the KSNR of new grids from these new general grid
designs was more than 35% to ten times higher than that in the literature.
The Tp (approximately equal to 1) and KSNR of all new grid designs implied that new grids with
these new grid designs would have the benefit of improved image diagnostic quality, regardless of
anatomy thicknesses, either in mammography or general radiography. Using grids built with these
new designs, an effective ionising radiation reduction management strategy in the implementation
of the ALARA principle can be achieved. This means that the lowest radiation exposure can be
achieved using such new grids to produce x-ray images without compromising image diagnostic
quality.
New grids from these new designs require further investigation of possible manufacturing defects
and implementation issues. This investigation can be performed through experimental evaluation
186
of a prototype of these new grid designs. Experimental evaluation of a prototype grid will help not
only verify any problems but will also refine these designs.
As these grid designs were evaluated under mammographic and general radiographic conditions,
these designs’ application to cone-beam computed tomography (CT) or fluoroscopy need further
investigation. Current grids were not found to improve image diagnostic quality of cone-beam or
flat-panel CT. As these new grids reduce a negligible amount of primary radiation, they should
improve image diagnostic quality and reduce patient radiation exposures in cone-beam or flat-
panel CT. The investigation of these new grid designs in cone-beam or flat-panel CT can be
undertaken using similar methods to that have been used in this project.
Clinical evaluation of these new grids from the new designs is not within the scope of this project.
The work undertaken in this project has not evaluated these new grids for improvements in
sensitivity and specificity of x-ray examinations. The new grids from the new designs have a
smaller Ts than that from contemporary grids. The new grids would improve sensitivity and
specificity of x-ray examinations much more than these contemporary grids. In addition, further
reduction of scatter radiation can be achieved without increasing patient radiation exposures by
using the new grid with a high grid ratio. Clinical evaluation of these new grids can be undertaken
on an individual x-ray examination basis, tailored to the anatomical region under investigation.
8.2. Summary
On commercial and clinical acceptance of these new grids using these new designs, image
diagnostic quality can be improved with reduced patient radiation exposure. As the operation of
these new grids will not be different from contemporary grids, staff training for using these grids
187
These new grids from the new designs are expected to produce the highest image diagnostic quality.
The application of these new grids should ultimately not only improve image diagnosis but also
188
Reference
ALZIMAMI, K., SASSI, S., ALKHORAYEF, M., BRITTEN, A. J. & SPYROU, M. M. 2009. Optimisation
of computed radiography systems for chest imaging. Nuclear Instruments and Methods in Physics
Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 600, 513-
518.
ATTIX, F. H. 2008. Introduction to radiological physics and radiation dosimetry, John Wiley & Sons.
BADAL, A. & BADANO, A. 2009. Accelerating Monte Carlo simulations of photon transport in a
voxelized geometry using a massively parallel graphics processing unit. Medical Physics, 36, 4878-
80.
BAKER, J. E., MOULDER, J. E. & HOPEWELL, J. W. 2011. Radiation as a risk factor for cardiovascular
disease. Antioxidants & Redox Signaling, 15, 1945-1956.
BARNES, G. T., CLEARE, H. M. & BREZOVICH, I. A. 1976. Reduction of scatter in diagnostic radiology
by means of a scanning multiple slit assembly. Radiology, 120, 691-4.
BARNES, G. T., WU, X. & WAGNER, A. J. 1993. Scanning slit mammography. Medical Progress
through Technology, 19, 7-12.
BELL, N., ERSKINE, M. & WARREN-FORWARD, H. 2003. Lateral cervical spine examinations: an
evaluation of dose for grid and non-grid techniques. Radiography, 9, 43-52.
BERNHARDT, P., MERTELMEIER, T. & HOHEISEL, M. 2006. X-ray spectrum optimization of full-
field digital mammography: simulation and phantom study. Medical Physics, 33, 4337-49.
BERNSTEIN, H., MUNTZ, E. P., SCHRECKENDGUST, J., KLEIN, D. J. & LEE, K. 1983. A detailed
experimental and theoretical comparison of the angular and energy dependencies of grid
transmission. Medical Physics, 10, 218-23.
BOICE JR, J. D., PRESTON, D., DAVIS, F. G. & MONSON, R. R. 1991. Frequent chest X-ray fluoroscopy
and breast cancer incidence among tuberculosis patients in Massachusetts. Radiation research, 125,
214-222.
BONENKAMP, J. G. & HONDIUS BOLDINGH, W. 1959a. Quality and choice of Potter Bucky grids. I.
A new method for the unambiguous determination of the quality of a grid. Acta Radiologica, 51,
479-89.
189
BONENKAMP, J. G. & HONDIUS BOLDINGH, W. 1959b. Quality and choice of Potter Bucky grids. III.
The choice of Bucky grid. Acta Radiologica, 52, 241-53.
BOONE, J. M. & COOPER III, V. N. 2000. Scatter/primary in mammography: Monte Carlo validation.
Medical Physics, 27, 1818-31.
BOONE, J. M., FEWELL, T. R. & JENNINGS, R. J. 1997. Molybdenum, rhodium, and tungsten anode
spectral models using interpolating polynomials with application to mammography. Medical
Physics, 24, 1863-1874.
BOONE, J. M., LINDFORS, K. K., COOPER III, V. N. & SEIBERT, J. A. 2000. Scatter/primary in
mammography: comprehensive results. Medical Physics, 27, 2408-16.
BOONE, J. M. & SEIBERT, J. A. 1988. Monte Carlo simulation of the scattered radiation distribution in
diagnostic radiology. Medical physics, 15, 713-720.
BOREMAN, G. D. 2001. Modulation transfer function in optical and electro-optical systems, SPIE Press
Bellingham, WA.
BORNEFALK, H., XU, C., SVENSSON, C. & DANIELSSON, M. 2010. Design considerations to
overcome cross talk in a photon counting silicon strip detector for computed tomography. Nuclear
Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors
and Associated Equipment, 621, 371-378.
BUCKY, G. 1913. A Grating-Diaphragm to Cut off Secondary Rays from the Object. Archives of the
Roentgen Ray, 18, 6-9.
BURGESS, A. E. 1999. The Rose model, revisited. Journal of the Optical Society of America A, 16, 633-
646.
BUSHBERG, J. T., SEIBERT, J. A., EDWIN M. LEIDHOLDT, J. & BOONE, J. M. 2011. The essential
physics of medical imaging, Lippincott Williams & Wilkins.
BUSHONG, S. C. 2013. Radiologic science for technologists: physics, biology, and protection, Elsevier
Health Sciences.
CARDOSO, S. C., GONÇALVES, O. D., SCHECHTER, H. & EICHLER, J. 2003. Modelling the elastic
scattering in diagnostic radiology: the importance of structure form factors. Physics in Medicine
and Biology, 48, 1907.
CARLTON, R. R., ADLER, A. M. & FRANK, E. D. 2012. Principles of radiographic imaging : an art
and a science, Clifton Park, N.Y., Thomson Delmar Learning.
190
CARTON, A. K., ACCIAVATTI, R., KUO, J. & MAIDMENT, A. D. 2009. The effect of scatter and glare
on image quality in contrast-enhanced breast imaging using an a-Si/CsI(TI) full-field flat panel
detector. Medical Physics, 36, 920-8.
CHAN, H. P. & DOI, K. 1982. Investigation of the performance of antiscatter grids: Monte Carlo simulation
studies. Physics in medicine and biology, 27, 785-803.
CHAN, H. P., FRANK, P. H., DOI, K., IIDA, N. & HIGASHIDA, Y. 1985. Ultra-high-strip-density
radiographic grids: a new antiscatter technique for mammography. Radiology, 154, 807-15.
CHANTLER, C. 2000. Detailed tabulation of atomic form factors, photoelectric absorption and scattering
cross section, and mass attenuation coefficients in the vicinity of absorption edges in the soft X-ray
(Z= 30–36, Z= 60–89, E= 0.1 keV–10 keV), addressing convergence issues of earlier work. Journal
of Physical and Chemical Reference Data, 29, 597-1056.
CHEN, H. 2016. PhD Disertation: Characterization and Optimization of Silicon-strip Detectors for
Mammography and Computed Tomography. KTH VETENSKAP OCH KONST.
CHEN, H., DANIELSSON, M., XU, C. & CEDERSTRÖM, B. 2015. On image quality metrics and the
usefulness of grids in digital mammography. Journal of Medical Imaging, 2, 013501-013507.
COMPTON, A. H. 1923. A quantum theory of the scattering of X-rays by light elements. Physical review,
21, 483.
CULLEN, D. E., HUBBELL, J. H. & KISSEL, L. 1997. EPDL97: the evaluated photon data library,’97
version. UCRL-50400, 6, 1-28.
CUNHA, D. M., TOMAL, A. & POLETTI, M. E. 2010. Evaluation of scatter-to-primary ratio, grid
performance and normalized average glandular dose in mammography by Monte Carlo simulation
including interference and energy broadening effects. Physics in medicine and biology, 55, 4335-
59.
CUNHA, D. M., TOMAL, A. & POLETTI, M. E. 2012. Optimization of x-ray spectra in digital
mammography through Monte Carlo simulations. Physics in medicine and biology, 57, 1919-35.
DANCE, D. R. & DAY, G. J. 1984. The computation of scatter in mammography by Monte Carlo methods.
Physics in Medicine and Biology, 29, 237-247.
DANCE, D. R., STRUDLEY, C. J., YOUNG, K. C., ODUKO, J. M., WHELEHAN, P. J. &
MUNGUTROY, E. L. 2012. Comparison of breast doses for digital tomosynthesis estimated from
patient exposures and using PMMA breast phantoms. Breast Imaging. Springer.
DANCE, D. R., THILANDER, A. K., SANDBORG, M., SKINNER, C. L., CASTELLANO, I. A. &
CARLSSON, G. A. 2000. Influence of anode/filter material and tube potential on contrast, signal-
to-noise ratio and average absorbed dose in mammography: a Monte Carlo study. British Journal
of Radiology, 73, 1056-67.
191
DAVIDSON, R. A. 2006. Radiographic contrast-enhancement masks in digital radiography. PhD doctoral
thesis, The University of Sydney.
DAY, G. J. & DANCE, D. R. 1983. X-ray transmission formula for antiscatter grids. Physics in medicine
and biology, 28, 1429-33.
DICK, C. E. & MOTZ, J. W. 1978. New method for the experimental evaluation of x-ray grids. Medical
Physics, 5, 133-40.
DICK, C. E., SOARES, C. G. & MOTZ, J. W. 1978. X-ray scatter data for diagnostic radiology. Physics
in medicine and biology, 23, 1076.
DOODY, M. M., LONSTEIN, J. E., STOVALL, M., HACKER, D. G., LUCKYANOV, N. & LAND, C.
E. 2000. Breast cancer mortality after diagnostic radiography: findings from the US Scoliosis
Cohort Study. Spine, 25, 2052-2063.
DORES, G. M., METAYER, C., CURTIS, R. E., LYNCH, C. F., CLARKE, E. A., GLIMELIUS, B.,
STORM, H., PUKKALA, E., VAN LEEUWEN, F. E., HOLOWATY, E. J., ANDERSSON, M.,
WIKLUND, T., JOENSUU, T., VAN’T VEER, M. B., STOVALL, M., GOSPODAROWICZ, M.
& TRAVIS, L. B. 2002. Second Malignant Neoplasms Among Long-Term Survivors of Hodgkin’s
Disease: A Population-Based Evaluation Over 25 Years. Journal of Clinical Oncology, 20, 3484-
3494.
DURBIN, J. & WATSON, G. S. 1951. Testing for serial correlation in least squares regression. II.
Biometrika, 38, 159-178.
ECKHARDT, R. 1987. Stan ulam, john von neumann, and the monte carlo method. Los Alamos Science,
15, 30.
EIDEMÜLLER, M., HOLMBERG, E., JACOB, P., LUNDELL, M. & KARLSSON, P. 2009. Breast cancer
risk among Swedish hemangioma patients and possible consequences of radiation-induced
genomic instability. Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis,
669, 48-55.
EIDEMÜLLER, M., HOLMBERG, E., JACOB, P., LUNDELL, M. & KARLSSON, P. 2011. Breast cancer
risk after radiation treatment at infancy: potential consequences of radiation-induced genomic
instability. Radiation protection dosimetry, 143, 375-379.
EVANS, S. H., BRADLEY, D. A., DANCE, D. R., BATEMAN, J. & JONES, C. H. 1991. Measurement
of small-angle photon scattering for some breast tissues and tissue substitute materials. Physics in
medicine and biology, 36, 7.
FETTERLY, K. A. & HANGIANDREOU, N. J. 2001. Effects of x-ray spectra on the DQE of a computed
radiography system. Medical Physics, 28, 241-249.
192
FETTERLY, K. A. & SCHUELER, B. A. 2007. Experimental evaluation of fiber-interspaced antiscatter
grids for large patient imaging with digital x-ray systems. Physics in medicine and biology, 52,
4863-80.
FLOYD, C. E., LO, J. Y., CHOTAS, H. G. & RAVIN, C. E. 1991. Quantitative scatter measurement in
digital radiography using a photostimulable phosphor imaging system. Medical Physics, 18, 408-
413.
FLYNN, M. J. & SAMEI, E. 1999. Experimental comparison of noise and resolution for 2k and 4k storage
phosphor radiography systems. Medical Physics, 26, 1612-23.
GEIJER, H., BECKMAN, K.-W., ANDERSSON, T. & PERSLIDEN, J. 2001. Image quality vs radiation
dose for a flat-panel amorphous silicon detector: a phantom study. European Radiology, 11, 1704-
1709.
GRAHAM, D. T., CLOKE, P. J. & VOSPER, M. 2003. Principles of radiological physics, Sydney,
Churchill Livingstone.
HALL, E. J. 2006. Radiobiology for the Radiologist, Lippincott Williams & Wilkins.
HALL, E. J. & GIACCIA, A. J. 2012. Radiobiology for the Radiologist, Lippincott Williams & Wilkins.
HENDEE, W. R. & RITENOUR, E. R. 2002. Medical imaging physics, New York, wiley-Liss, Inc.
HERRNSDORF, L. & PETERSSON, H. 2016. BEAM QUALITY CORRECTION FACTORS FOR KAP
METERS FOR LIGHTLY AND HEAVILY FILTERED X-RAY BEAMS. Radiation Protection
Dosimetry.
HONDIUS BOLDINGH, W. 1961. Quality and Choice of Potter Bucky Grids Parts IV and V. Acta
Radiologica, 55, 225-235.
HUBBELL, J. H. 2006. Review and history of photon cross section calculations. Physics in Medicine and
Biology, 51, R245-R262.
193
HUBBELL, J. H. & SELTZER, S. M. 1995. Tables of X-ray mass attenuation coefficients and mass energy-
absorption coefficients 1 keV to 20 MeV for elements Z= 1 to 92 and 48 additional substances of
dosimetric interest. National Inst. of Standards and Technology-PL, Gaithersburg, MD (United
States). Ionizing Radiation Div.
HUBBELL, J. H., VEIGELE, W. J., BRIGGS, E. A., BROWN, R. T., CROMER, D. T. & HOWERTON,
R. J. 1975. Atomic form factors, incoherent scattering functions, and photon scattering cross
sections. Journal of physical and chemical reference data, 4, 471-538.
HUDA, W., SAJEWICZ, A. M., OGDEN, K. M. & DANCE, D. R. 2003. Experimental investigation of
the dose and image quality characteristics of a digital mammography imaging system. Medical
Physics, 30, 442-8.
ICRU 1989. Tissue substitutes in radiation dosimetry and measurement. Report 44 of the International
Commission on Radiation Units and Measurements (Bethesda, MD).
194
INTERNATIONAL ELECTROTECHNICAL COMMISSION 2001. Medical Diagnostic X-ray equipment:
Characteristics of general purpose and mammographic antis-catter grids. Geneva: IEC.
JAFFE, C. & WEBSTER, E. W. 1975. Radiographic contrast improvement by means of slit radiography.
Radiology, 116, 631-5.
JANG, D. Y., KIM, D. I., PARK, C. J., CHOE, B. Y., SUH, T. S., LEE, H. K., JUNG, N. K., KIM, J. S.,
SEO, S. W. & SEONG, W. 2008. Performance evaluation of carbon-interspaced anti-scatter grids
in mammography: Empirical formula and Monte Carlo simulation studies. Journal of the Korean
Physical Society, 53, 863-867.
JING, Z., HUDA, W. & WALKER, J. K. 1998. Scattered radiation in scanning slot mammography. Medical
Physics, 25, 1111-7.
JOHNS, P. C. & YAFFE, M. 1982. Scattered radiation in fan beam imaging systems. Medical Physics, 9,
231-9.
KADHIM, M., SALOMAA, S., WRIGHT, E., HILDEBRANDT, G., BELYAKOV, O. V., PRISE, K. M.
& LITTLE, M. P. 2013. Non-targeted effects of ionising radiation—Implications for low dose risk.
Mutation Research, 752, 84-98.
KALENDER, W. 1981. Monte Carlo calculations of x-ray scatter data for diagnostic radiology. Physics in
medicine and biology, 26, 835-849.
KALENDER, W. A. 1982. Calculation of x-ray grid characteristics by Monte Carlo methods. Physics in
medicine and biology, 27, 353-61.
KASAP, S., FREY, J. B., BELEV, G., TOUSIGNANT, O., MANI, H., GREENSPAN, J., LAPERRIERE,
L., BUBON, O., REZNIK, A. & DECRESCENZO, G. 2011. Amorphous and polycrystalline
photoconductors for direct conversion flat panel x-ray image sensors. Sensors, 11, 5112-5157.
KAWRAKOW, I. 2000. Accurate condensed history Monte Carlo simulation of electron transport. II.
Application to ion chamber response simulations. Medical physics, 27, 499-513.
195
KHONG, P. L., RINGERTZ, H., DONOGHUE, V., FRUSH, D., REHANI, M., APPELGATE, K. &
SANCHEZ, R. 2013. ICRP PUBLICATION 121: Radiological Protection in Paediatric Diagnostic
and Interventional Radiology. Annals of the ICRP.
KIM, K. Y., OH, J. E., LEE, S. Y., CHOI, S. I., CHO, H. S., KIM, J. S., CHUNG, N. G., KIM, J. W. &
KIM, J. G. 2007. Performance evaluate of carbon-interspaced and aluminum-interspaced
antiscatter grids based upon the IEC standard fixture. Nuclear Science Symposium Conference
Record, 2007. NSS '07. IEEE.
KIM, S., YOSHIZUMI, T. T., TONCHEVA, G., FRUSH, D. P. & YIN, F.-F. 2010. Estimation of absorbed
doses from paediatric cone-beam CT scans: MOSFET measurements and Monte Carlo simulations.
Radiation Protection Dosimetry, 138, 257-263.
KOEDOODER, K. & VENEMA, H. W. 1986. Filter materials for dose reduction in screen-film
radiography. Physics in medicine and biology, 31, 585-600.
KROL, A., BASSANO, D. A., CHAMBERLAIN, C. C. & PRASAD, S. C. 1996. Scatter reduction in
mammography with air gap. Medical Physics, 23, 1263-70.
KYRIAKOU, Y. & KALENDER, W. 2007. Efficiency of antiscatter grids for flat-detector CT. Physics in
medicine and biology, 52, 6275-93.
LANDAU, D. P. & BINDER, K. 2009. A guide to Monte Carlo simulations in statistical physics,
Cambridge university press.
LAZOS, D. & WILLIAMSON, J. F. 2010. Monte Carlo evaluation of scatter mitigation strategies in cone-
beam CT. Medical physics, 37, 5456-5470.
LITTLE, M. P. 2003. Risks associated with ionizing radiation. British Medical Bulletin, 68, 259-75.
LITTLE, M. P. 2010. Do non-targeted effects increase or decrease low dose risk in relation to the linear-
non-threshold (LNT) model? Mutation Research, 687, 17-27.
LITTLE, M. P., HOEL, D. G., MOLITOR, J., BOICE, J. D., WAKEFORD, R. & MUIRHEAD, C. R.
2008a. New Models for Evaluation of Radiation-Induced Lifetime Cancer Risk and its Uncertainty
Employed in the UNSCEAR 2006 Report. Radiation Research, 169, 660-676.
LITTLE, M. P., TAWN, E. J., TZOULAKI, I., WAKEFORD, R., HILDEBRANDT, G., PARIS, F., TAPIO,
S. & ELLIOTT, P. 2008b. A systematic review of epidemiological associations between low and
moderate doses of ionizing radiation and late cardiovascular effects, and their possible mechanisms.
Radiation Research, 169, 99-109.
LITTLE, M. P., WAKEFORD, R. & KENDALL, G. M. 2009a. Updated estimates of the proportion of
childhood leukaemia incidence in Great Britain that may be caused by natural background ionising
radiation. Journal of Radiological Protection, 29, 467.
196
LITTLE, M. P., WAKEFORD, R., TAWN, E. J., BOUFFLER, S. D. & DE GONZALEZ, A. B. 2009b.
Risks associated with low doses and low dose rates of ionizing radiation: Why linearity may be
(almost) the best we can do. Radiology, 251, 6-12.
MALUSEK, A., SANDBORG, M. & CARLSSON, G. A. Simulation of scatter in cone beam CT: Effects
on projection image quality. In: YAFFE, M. J. & ANTONUK, L. E., eds. Procedings of SPIE Vol
5030, 2003. Physics of Medical Imaging, 740-751.
MALUSEK, A., SANDBORG, M. & CARLSSON, G. A. 2008. CTmod—A toolkit for Monte Carlo
simulation of projections including scatter in computed tomography. Computer methods and
programs in biomedicine, 90, 167-178.
MARANDI, A., LEINDECKER, N. C., VODOPYANOV, K. L. & BYER, R. L. 2012. All-optical quantum
random bit generation from intrinsically binary phase of parametric oscillators. Optics express, 20,
19322-19330.
MENTRUP, D., JOCKEL, S., MENSER, B., NEITZEL, U., D. E. HAMBURG & EINDHOVEN, N. L.
2014. Grid-like contrast restoration for non-grid chest radiographs by software-based scatter
correction.
MORGAN, R. H. 1946a. An analysis of the physical factors controlling the diagnostic quality of roentgen
images: Part IV. Contrast and the film contrast factor. The American journal of roentgenology &
radium therapy, 55, 627-33.
MORGAN, R. H. 1946b. An analysis of the physical factors controlling the diagnostic quality of Roentgen
images: Part III. Contrast and the intensity distribution function of a roentgen image. The American
journal of roentgenology & radium therapy, 55, 67-89.
MORGAN, R. H. 1949. An analysis of the physical factors controlling the diagnostic quality of roentgen
images; unsharpness. The American journal of roentgenology & radium therapy, 62, 870-80.
MORIN, L. R. M. 1982. Molecular form factors and photon coherent scattering cross sections of water.
Journal of Physical and Chemical Reference Data, 11, 1091-1098.
NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY 2016. XCOM: Photon Cross Sections
Database. http://www.nist.gov/pml/data/xcom/index.cfm.
NATIONAL RESEARCH COUNCIL 2006. Health Risks from Exposure to Low Levels of Ionizing
Radiation BEIR VII Phase 2, Washington, DC, National Research Council of the National
Academies.
197
NEITZEL, U. 1992. Grids or air gaps for scatter reduction in digital radiography: a model calculation.
Medical Physics, 19, 475-481.
NIKLASON, L. T., SORENSON, J. A. & NELSON, J. A. 1981. Scattered Radiation in Chest Radiography.
Medical Physics, 8, 677-681.
NIKOLOPOULOS, D., LINARDATOS, D., VALAIS, I., MICHAIL, C., DAVID, S., GONIAS, P.,
BERTSEKAS, N., CAVOURAS, D., LOUIZI, A. & KANDARAKIS, I. 2007. Monte Carlo
validation in the diagnostic radiology range. Nuclear Instruments and Methods in Physics Research
Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 571, 267-269.
OLEARY, D. D., GRANT, M. T. & RAINFORD, D. L. 2011. Image quality and compression force the
forgotten link in optimisaion of digital mammography. UCD School of Medicine and Medical
Science, Dublin, Ireland: health research board.
PERSLIDEN, J. & CARLSSON, G. A. 1997. Scatter rejection by air gaps in diagnostic radiology.
Calculations using a Monte Carlo collision density method and consideration of molecular
interference in coherent scattering. Physics in medicine and biology, 42, 155.
PICANO, E., VANO, E., DOMENICI, L., BOTTAI, M. & THIERRY-CHEF, I. 2012. Cancer and non-
cancer brain and eye effects of chronic low-dose ionizing radiation exposure. BMC Cancer, 12,
157.
PODGORŠAK, E. B. 2006. Rutherford—Bohr Atomic Model. Radiation Physics for Medical Physicists,
42-85.
POLETTI, M., GONÇALVES, O. & MAZZARO, I. 2002. X-ray scattering from human breast tissues and
breast-equivalent materials. Physics in medicine and biology, 47, 47.
POLUDNIOWSKI, G., EVANS, P. & WEBB, S. 2009a. Rayleigh scatter in kilovoltage x-ray imaging: is
the independent atom approximation good enough? Physics in medicine and biology, 54, 6931.
POLUDNIOWSKI, G., LANDRY, G., DEBLOIS, F., EVANS, P. M. & VERHAEGEN, F. 2009b.
SpekCalc: a program to calculate photon spectra from tungsten anode x-ray tubes. Physics in
medicine and biology, 54, N433-8.
PRESS, W. H. & TEUKOLSKY, S. A. 1992. Portable random number generators. Computers in physics,
6, 522-524.
PRESTON, D., RON, E., TOKUOKA, S., FUNAMOTO, S., NISHI, N., SODA, M., MABUCHI, K. &
KODAMA, K. 2007. Solid cancer incidence in atomic bomb survivors: 1958-1998. Radiation
research, 168, 1-64.
198
PRESTON, D. L., KUSUMI, S., TOMONAGA, M., IZUMI, S., RON, E., KURAMOTO, A., KAMADA,
N., DOHY, H., MATSUI, T. & NONAKA, H. 1994. Cancer incidence in atomic bomb survivors.
Part III: Leukemia, lymphoma and multiple myeloma, 1950-1987. Radiation research, 137, S68-
S97.
PRESTON, D. L., SHIMIZU, Y., PIERCE, D. A., SUYAMA, A. & MABUCHI, K. 2003. Studies of
mortality of atomic bomb survivors. Report 13: Solid cancer and noncancer disease mortality:
1950-1997. Radiation research, 160, 381-407.
RENGER, B., BRIESKORN, C., TOTH, V., MENTRUP, D., JOCKEL, S., LOHÖFER, F., SCHWARZ,
M., RUMMENY, E. J. & NOËL, P. B. 2016. EVALUATION OF DOSE REDUCTION
POTENTIALS OF A NOVEL SCATTER CORRECTION SOFTWARE FOR BEDSIDE CHEST
X-RAY IMAGING. Radiation protection dosimetry, 169, 60-67.
ROSE, A. 1948. The sensitivity performance of the human eye on an absolute scale. JOSA, 38, 196-208.
ROYLE, G. J. & SPELLER, R. D. 1995. Quantitative x-ray diffraction analysis of bone and marrow
volumes in excised femoral head samples. Physics in medicine and biology, 40, 1487.
SALVAT, F. PENELOPE-2015, A code system for Monte Carlo simulation of electron and phonon
transport. Workshop proceedings, 2015 Barcelona, Spain. OECD Nuclear Energy Agency.
SAUNDERS JR, R. S. & SAMEI, E. A Monte Carlo investigation on the impact of scattered radiation on
mammographic resolution and noise. Proc. SPIE, Medical Imaging: Physics of Medical Imaging,
2006. 61423A-7A.
SHARMA, S., DURAND, R. M. & GUR-ARIE, O. 1981. Identification and analysis of moderator variables.
Journal of marketing research, 291-300.
SHUPING, R. E. & JUDY, P. F. 1977. Energy absorbed in calcium tungstate x‐ray screens. Medical
Physics, 4, 239-243.
SISNIEGA, A., ZBIJEWSKI, W., BADAL, A., KYPRIANOU, I. S., STAYMAN, J. W., VAQUERO, J. J.
& SIEWERDSEN, J. H. 2013. Monte Carlo study of the effects of system geometry and antiscatter
grids on cone-beam CT scatter distributions. Medical Physics, 40, 051915.
199
SORENSON, J. A. & FLOCH, J. 1985. Scatter rejection by air gaps: an empirical model. Medical Physics,
12, 308-16.
STAR-LACK, J., SUN, M., KAESTNER, A., HASSANEIN, R., VIRSHUP, G., BERKUS, T. &
OELHAFEN, M. Efficient scatter correction using asymmetric kernels. SPIE Medical Imaging,
2009. International Society for Optics and Photonics, 72581Z-72581Z-12.
STRID, K. G. 1976. Analysis of secondary screening with special reference to grids for abdominal
radiography. Acta Radiologica, 351, 1-113.
STURM, R. E. & MORGAN, R. H. 1949. Screen intensification systems and their limitations. American
Journal of Roentgenology and Radium Therapy, 62, 617-634.
SUN, M. & STAR-LACK, J. M. 2010. Improved scatter correction using adaptive scatter kernel
superposition. Physics in medicine and biology, 55, 6695-6720.
TAKAHIRO KAWAMURA, SATOSHI NAITO, KAYO OKANO & YAMADA, M. 2015. Improvement
in Image Quality and Workflow of X-Ray Examinations Using a New Image Processing Method,
"Virtual Grid Technology". FUJIFILM RESEARCH & DEVELOPMENT (No. 60-2015).
TANAKA, N., NAKA, K., SAITO, A., MORISHITA, J., TOYOFUKU, F., OHKI, M. & HIGASHIDA, Y.
2013. Investigation of optimum anti-scatter grid selection for digital radiography: physical imaging
properties and detectability of low-contrast signals. Radiological physics and technology, 6, 54-60.
TARTARI, A., TAIBI, A., BONIFAZZI, C. & BARALDI, C. 2002. Updating of form factor tabulations
for coherent scattering of photons in tissues. Physics in Medicine and Biology, 47, 163.
TER-POGOSSIAN, M. M., PHELPS, M. E., HOFFMAN, E. J. & EICHUNG, J. O. 1974. The Extraction
of the Yet Unused Wealth of Information in Diagnostic Radiology. Radiology, 113, 515-520.
TOMAL, A., CUNHA, D. M. & POLETTI, M. E. 2013. Optimal X-Ray Spectra Selection in Digital
Mammography: A Semi-Analytical Study. Nuclear Science, IEEE Transactions on, 60, 728-734.
TORTORA, G. J. & DERRICKSON, B. H. 2008. Principles of anatomy and physiology, John Wiley &
Sons.
200
TRAVIS, L. B., GOSPODAROWICZ, M., CURTIS, R. E., CLARKE, E. A., ANDERSSON, M.,
GLIMELIUS, B., JOENSUU, T., LYNCH, C. F., VAN LEEUWEN, F. E. & HOLOWATY, E.
2002. Lung cancer following chemotherapy and radiotherapy for Hodgkin's disease. Journal of the
National Cancer Institute, 94, 182-192.
TROMANS, C., DIFFEY, J. & BRADY, S. 2010. Investigating the Replacement of the Physical Anti-
scatter Grid with Digital Image Processing. In: MARTÍ, J., OLIVER, A., FREIXENET, J. &
MARTÍ, R. (eds.) Digital Mammography. Springer Berlin Heidelberg.
TUBIANA, M., AURENGO, A., AVERBECK, D. & MASSE, R. 2006. The debate on the use of linear no
threshold for assessing the effects of low doses. Journal of Radiological Protection, 26, 317-24.
ULLMAN, G., SANDBORG, M., DANCE, D. R., HUNT, R. A. & CARLSSON, G. A. 2006. Towards
optimization in digital chest radiography using Monte Carlo modelling. Physics in medicine and
biology, 51, 2729-2743.
WEBBER, R. L., YOUMANS, H. D. & NAGEL, R. N. 1977. A nonlinear model for predicting radiographic
contrast. Oral Surgery, Oral Medicine, Oral Pathology, 43, 798-811.
WILSEY, R. B. 1921. The intensity of scattered x-rays in radiography. Am. J. Roentgenol, 8, 328.
YAFFE, M. & ROWLANDS, J. 1997. X-ray detectors for digital radiography. Physics in Medicine and
Biology, 42, 1.
YAN, H., MOU, X., TANG, S., XU, Q. & ZANKL, M. 2010. Projection correlation based view
interpolation for cone beam CT: primary fluence restoration in scatter measurement with a moving
beam stop array. Physics in medicine and biology, 55, 6353-75.
YU, C., LIU, B., O'CONNOR, J. M., DIDIER, C. S. & GLICK, S. J. 2009. Characterization of scatter in
cone-beam CT breast imaging: Comparison of experimental measurements and Monte Carlo
simulation. Medical Physics, 36, 857-869.
ZBIJEWSKI, W., SISNIEGA, A., VAQUERO, J. J., MUHIT, A., PACKARD, N., SENN, R., YANG, D.,
YORKSTON, J., CARRINO, J. A. & SIEWERDSEN, J. H. Dose and scatter characteristics of a
novel cone beam CT system for musculoskeletal extremities. SPIE Medical Imaging, vol 8313,
2012.
ZHAO, W., RISTIC, G. & ROWLANDS, J. 2004. X-ray imaging performance of structured cesium iodide
scintillators. Medical physics, 31, 2594-2605.
ZHOU, A., YIN, Y., WHITE, G. L. & DAVIDSON, R. 2016. A new method for radiation transmission of
anti-scatter grids. Biomedical Physics & Engineering Express, 2, 055011.
201
Appendix
Publications arising from this work are listed below. They can be found from their DOI links.
Zhou, Abel, Yuming Yin, Graeme L. White, and Rob Davidson. 2016. "A new method for radiation
transmission of anti-scatter grids." Biomedical Physics & Engineering Express 2 (5):055011. doi:
https://doi.org/10.1088/2057-1976/2/5/055011.
Zhou, Abel, Graeme L. White, and Rob Davidson. 2018. "Validation of a Monte Carlo code
system for grid evaluation with interference effect on Rayleigh scattering." Physics in Medicine &
Biology 63 (3). doi: https://doi.org/10.1088/1361-6560/aaa44b.
203