Home Search Collections Journals About Contact us My IOPscience

Monte Carlo simulation of x-ray spectra in diagnostic radiology and mammography using

MCNP4C

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2004 Phys. Med. Biol. 49 4897

(http://iopscience.iop.org/0031-9155/49/21/004)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 64.141.84.23
This content was downloaded on 24/11/2014 at 21:44

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 49 (2004) 4897–4917 PII: S0031-9155(04)77842-2

Monte Carlo simulation of x-ray spectra in diagnostic

radiology and mammography using MCNP4C
M R Ay1,2, M Shahriari3, S Sarkar4, M Adib5 and H Zaidi2
1 Department of Physics and Nuclear Sciences, AmirKabir University of Technology, Tehran, Iran
2 Division of Nuclear Medicine, Geneva University Hospital, 1211 Geneva, Switzerland
3 Department of Nuclear Engineering, Shahid Beheshti University, Tehran, Iran
4 Department of Medical Physics, Tehran University of Medical Science, Tehran, Iran
5 TPP Co., GE Medical Systems, Iran Authorized Distributor, Tehran, Iran

E-mail: farshid.ay@tppgems.com

Received 16 March 2004

Published 8 October 2004
Online at stacks.iop.org/PMB/49/4897
doi:10.1088/0031-9155/49/21/004

Abstract
The general purpose Monte Carlo N-particle radiation transport computer code
(MCNP4C) was used for the simulation of x-ray spectra in diagnostic radiology
and mammography. The electrons were transported until they slow down and
stop in the target. Both bremsstrahlung and characteristic x-ray production were
considered in this work. We focus on the simulation of various target/filter
combinations to investigate the effect of tube voltage, target material and
filter thickness on x-ray spectra in the diagnostic radiology and mammography
energy ranges. The simulated x-ray spectra were compared with experimental
measurements and spectra calculated by IPEM report number 78. In addition,
the anode heel effect and off-axis x-ray spectra were assessed for different
anode angles and target materials and the results were compared with EGS4-
based Monte Carlo simulations and measured data. Quantitative evaluation of
the differences between our Monte Carlo simulated and comparison spectra
was performed using student’s t-test statistical analysis. Generally, there
is a good agreement between the simulated x-ray and comparison spectra,
although there are systematic differences between the simulated and reference
spectra especially in the K-characteristic x-rays intensity. Nevertheless, no
statistically significant differences have been observed between IPEM spectra
and the simulated spectra. It has been shown that the difference between MCNP
simulated spectra and IPEM spectra in the low energy range is the result of the
overestimation of characteristic photons following the normalization procedure.
The transmission curves produced by MCNP4C have good agreement with the
IPEM report especially for tube voltages of 50 kV and 80 kV. The systematic
discrepancy for higher tube voltages is the result of systematic differences
between the corresponding spectra.
(Some figures in this article are in colour only in the electronic version)
0031-9155/04/214897+21$30.00 © 2004 IOP Publishing Ltd Printed in the UK 4897
4898 M R Ay et al

1. Introduction

Computer simulation of x-ray spectra is one of the most important tools for investigation
of patient dose and image quality in diagnostic radiology imaging systems. Early attempts
at predicting diagnostic x-ray spectra were undertaken by Kramers (1923). This pioneering
work was sustained by several investigators and many research groups are still trying to
find an accurate method for computer simulation of x-ray spectra owing to the fact that
experimental measurement of x-ray spectra requires special equipment which is available
only in a limited number of laboratories (Fewell and Shuping 1978, Fewell et al 1981,
Laitano et al 1991, Antonuk et al 1997, Dance et al 2000, Wilkinson et al 2001). Fewell
et al measured x-ray spectra with different target/filter combinations for over two decades
and have published several measured spectra (Fewell and Shuping 1977, 1978, Fewell et al
1981). Since experimental measurement of x-ray spectra is time consuming and remains
difficult, different methods for spectra prediction have been presented. These can be
divided into three categories: empirical models (Fewell and Shuping 1977, Boone and Seibert
1997, Boone et al 1997), semi-empirical models (Birch and Marshall 1979, Boone 1988,
Tucker et al 1991, Blough et al 1998) and Monte Carlo modelling (Kulkarni and Supe 1984,
Acosta et al 1998, Bhat et al 1999, Verhaegen et al 1999, Ng et al 2000, Ben Omrane et al
2003, Verhaegen and Castellano 2002). Although, purely empirical and semi-empirical
models remain the fastest methods for x-ray spectra prediction, the models proposed so far
still have limitations which prevent their adoption for a large range of applications (Ay et al
2004). In addition, most of these models have preset target/filter combinations and thus do not
allow investigation of newly developed material compositions on the quality of resulting x-ray
spectra. Sophisticated Monte Carlo modelling was adopted as an alternative to overcome
the limitations mentioned above. Nevertheless, the prediction of x-ray spectra using the
Monte Carlo method is computer intensive and time consuming compared to empirical and
semi-empirical models.
The use of the Monte Carlo method to simulate radiation transport has become the most
accurate means of predicting the x-ray spectra even in complex geometries owing to more
accurate physics modelling and incorporation of appropriate interaction cross section data
(Zaidi and Sgouros 2002). Moreover, the method tracks the evolution of all secondary particles
(and their descendants) generated by primary electrons. Using the Monte Carlo method, it
is possible to transport electrons and photons inside the target and filter to obtain detailed
information about the factors contributing to the production of the x-ray spectrum (Acosta et al
1998). For the purpose of Monte Carlo simulation of x-ray spectrum, some authors have used
self-written or in house developed computer codes (Kulkarni and Supe 1984, O’Meara et al
1998), while others have used public domain general-purpose Monte Carlo codes such as EGS4
(Bhat et al 1998, 1999, Ben Omrane et al 2003), MCNP (Verhaegen et al 1999, Mercier et al
2000) and ITS (Ng et al 2000).
In this work, we used MCNP4C running on Pentium-based PC to simulate the diagnostic
radiology and mammography x-ray tube with the aim of predicting the x-ray spectra with
different combinations of target/filter such as W/Al, Mo/Mo and Mo/Rh using various tube
voltages (between 50 and 140 kV in diagnostic radiology and 30 kV in mammography) and
different anode angles (between 6◦ and 18◦ ). The Be window and air between tube window
and measurement point were simulated. This paper addresses particularly some aspects not
sufficiently covered in previously published papers, namely the variation of the radiation
output across the x-ray beam, the anode heel effect and off-axis spectra for different target
angles both in diagnostic radiology and mammography. The main motivations behind the
choice of this code are its wide use by the medical physics community, wide acceptance
Monte Carlo simulation of x-ray spectra using MCNP4C 4899

anode ψ cathode

calculation points (A)

Be filter

additional filter

calculation points (B)

Figure 1. Geometry of the experimental set-up used for Monte Carlo simulation of x-ray spectra
and assessment of anode heel effect. The position of detectors for calculation of exposure around
the target (A) and anode heel (B) is also shown.

as an international standard for coupled particle transport having the best condensed history
electron physics package and remarkable tally capabilities in addition to a powerful reporting
system of statistical checks (Briesmeister 2000, Mercier et al 2000). The validity of MCNP4C
simulated data was checked by comparing the calculated spectra, transmission curves and
heel effect with the IPEM report number 78 (Cranley et al 1997), measured data (Fewell et al
1981, Bhat et al 1998, 1999, Pernieka et al 1997) and EGS4-based Monte Carlo simulations
(Bhat et al 1999), respectively.

2. Material and methods

2.1. The MCNP4C code

MCNP is a general-purpose Monte Carlo code that can be used for neutron, photon and
electron or coupled neutron/photon/electron transport (Briesmeister 2000). The code treats
an arbitrary three-dimensional configuration of materials in geometric cells bounded by first
and second degree surfaces and fourth degree elliptical tori. For photons transport, the
code takes into account incoherent and coherent scattering, the possibility of fluorescent
emission after photoelectric absorption and bremsstrahlung. The continuous slowing down
approximation energy loss model is used for electron transport. To follow an electron through
a significant energy loss, the MCNP4C code breaks the electron’s path into many steps. These
steps are chosen to be long enough to encompass many collisions (so that multiple scattering
theories are valid) but short enough so that the mean energy loss in any one step is small
(so that the approximations necessary for multiple scattering theories are satisfied). Except
for the energy loss and straggling calculation, the detailed simulation of the electron history
takes place in the sampling of the substeps. The Goudsmit–Saunderson theory is used to
sample from the distribution of angular deflections, so that the direction of the electron can
change at the end of each substep. For electron transport, MCNP addresses the sampling of
bremsstrahlung photons at each electron substep. The table of production probabilities is used
4900 M R Ay et al

2.00E-01
MCNP
1.80E-01 IPEM

1.60E-01

Normalized No./keV
1.40E-01

1.20E-01

1.00E-01

8.00E-02

6.00E-02

120 kV 140 kV
4.00E-02
100 kV
80 kV
2.00E-02

0.00E+00
0 20 40 60 80 100 120 140
(a) Energy (keV)

0.08
MCNP
IPEM
0.07

0.06
Normalized No./keV

0.05

0.04

0.03

0.02

0.01

0
0 20 40 60 80 100 120 140

(b) Energy (keV)

Figure 2. (a) Comparison of x-ray spectra produced by MCNP4C and IPEM report 78 for various
tube voltages. (b) Same as (a) for 140 kV tube voltage with 0.1 mm Cu additional filter.

to determine whether a bremsstrahlung photon will be created (Hughes 1997). To improve the
efficiency of electron and photon transport, two cards (PHYS:P and PHYS:E) are implemented
in MCNP for biasing some physical parameters such as production of secondary electrons by
photons (IDES), coherent scattering (NOCOH), bremsstrahlung angular distribution (IBAD)
and production of characteristic x-rays (XNUM). The default value for the latter (XNUM = 1)
results in the analog number of tracks being sampled. If XNUM > 0, the number of photons
produced is XNUM times the number that would be produced in the analog case, and a
corresponding weight adjustment is made. Setting XNUM to zero turns off the production of
x-ray photons by electrons.

2.2. Simulation of x-ray spectra using MCNP4C

The procedure of x-ray production consists of tracking a large number of electrons incident
on the target until they are absorbed or emerge from it, and calculating the number of
Monte Carlo simulation of x-ray spectra using MCNP4C 4901

1.00E-01
MCNP
Measured (Bhat et al 1998)

8.00E-02

Normalized No./keV
6.00E-02

4.00E-02

2.00E-02

0.00E+00
0 10 20 30 40 50 60
(a) Energy (keV)

6.00E-02
MCNP
Measured (Bhat et al 1998)
Measured (Fewell et al 1981)

5.00E-02
Normalized No./keV

4.00E-02

3.00E-02

2.00E-02

1.00E-02

0.00E+00
0 10 20 30 40 50 60 70 80 90
(b) Energy (keV)

Figure 3. Comparison of x-ray spectra for various tube voltages produced by MCNP4C and
measured data by Fewell et al (1981) and Bhat et al (1998). (a) 50 kV, (b) 80 kV, (c) 100 kV.

bremsstrahlung and characteristic photons produced by them during their travel within the
target (Kulkarni and Supe 1984). For simulation of x-ray spectra, MCNP4C was run in
photon and electron mode (mode: P, E) using default values for PHYS:P and PHYS:E cards
to enable full electron and photon transport (XNUM was modified in our experiments to
investigate the intensity of produced characteristic photons).
The procedure starts with definition of an electron source (defined in our experiments as
a point source) emitting the electrons with energy E within a solid angle ψ towards the target.
The focal spot size on the target can be adjusted by changing this angle. Focal spot sizes
of 0.3, 0.6 and 1.2 mm corresponding to the projection on the beam central axis of the large
diameter of the ellipse covered by solid angle ψ on the target were investigated. Although the
approximation of the focal spot shape by an ellipse is a rough estimation, it was considered
reliable for the assessment of the influence of focal size on heel effect.
We have considered a constant energy for emitted electrons in most of the cases, although
the effect of ripple in electrons energy was simulated in one experiment to study its effect on
4902 M R Ay et al

6.00E-02
MCNP
Measured (Bhat et al 1998)
Measured (Fewell et al 1981)

5.00E-02

4.00E-02
Normalized No./keV

3.00E-02

2.00E-02

1.00E-02

0.00E+00
0 20 40 60 80 100
(c) Energy (keV)

Figure 3. (Continued.)

1.80E-02
MCNP
IPEM
1.60E-02

Ripple=0%
1.40E-02
Ripple=10%
Ripple=20%
Normalized No./keV

1.20E-02
Ripple=30%

1.00E-02

8.00E-03

6.00E-03

4.00E-03

2.00E-03

0.00E+00
0 10 20 30 40 50 60 70 80 90
Energy (keV)

Figure 4. Comparison of x-ray spectra using various tube voltage ripple factors in 80 kV produced
by MCNP4C and IPEM report 78.

the calculated spectra. To simulate ripple in MCNP4C, we used the following formulation for
ripple factor and tube voltage (Boone and Seibert 1997):
kVmax − kVmin
Ripple factor (RF) = 100 × . (1)
kVmax
Thus, the energy of an electron before emission towards the target is sampled uniformly from
tube voltage waveform with different ripple values:


RF
kV (t) = kVmax 1 − (1 − |sin(ωt)|) (2)
100
where t is time. When the ripple is neglected (i.e. RF = 0), kVmax = kV .
Monte Carlo simulation of x-ray spectra using MCNP4C 4903

5.00E-01
MCNP
IPEM
I
4.50E-01

4.00E-01

3.50E-01

Normalized No./ keV

3.00E-01

2.50E-01

2.00E-01

1.50E-01

1.00E-01

5.00E-02

0.00E+00
0 5 10 15 20 25 30 35
(a) Energy (keV)

5.00E-01
MCNP
IPEM
4.50E-01

4.00E-01

3.50E-01
Normalized No./ keV

3.00E-01

2.50E-01

2.00E-01

1.50E-01

1.00E-01

5.00E-02

0.00E+00
0 5 10 15 20 25 30 35
(b) Energy (keV)

Figure 5. Comparison of 30 kV x-ray spectra produced by MCNP4C and IPEM report 78 for
(a) Mo target with 0.5 mm Be and 0.03 mm Mo filter, (b) Mo target with 0.5 mm Be and
0.025 mm Rh and (c) W target with 0.5 mm Be and 1.2 mm Al.

When the electrons strike the target, the code transports the electrons inside the target
material until they are stopped after losing their kinetic energy. During the electrons’ transport,
all bremsstrahlung and characteristic x-ray production is considered. The calculated spectrum
is then normalized to the total number of photons in the spectrum. The experimental set-up
used in our simulation was based on the RAD 60 x-ray tube (Varian Medical Systems, CA)
material composition and target dimensions (figure 1).
MCNP4C simulations without applying variance reduction techniques require an
unacceptably long time to produce statistically relevant results. Thus, a variance reduction
technique known as point detector (F5 tally), belonging to the class of partially-deterministic
variance reduction methods implemented in MCNP4C was used. In this method, the transport
of particles towards the detector is replaced by a deterministic estimate of potential contribution
to the detector (Briesmeister 2000). The point detector tally measures photon flux at a point
4904 M R Ay et al

5.00E-01
MCNP
IPEM
4.50E-01

4.00E-01

3.50E-01

Normalized No./ keV

3.00E-01

2.50E-01

2.00E-01

1.50E-01

1.00E-01

5.00E-02

0.00E+00
0 5 10 15 20 25 30 35
(c) Energy (keV)

Figure 5. (Continued.)

(unit is photons cm−2 or MeV cm−2), which is normalized to be per starting particle. We
used the unit photons cm−2 for spectra generation and MeV cm−2 for exposure assessment
in the part related to investigation of anode heel effect. The point detectors were arranged in
the calculation points A and B shown in figure 1 to calculate the exposure around the target
and study anode heel effect. Note that the arrangement of point detectors for the calculation
of exposure in the axis perpendicular to the anode–cathode axis is not shown. According to
the MCNP user manual (Briesmeister 2000), an uncertainty of less than 5% is required for
point detector tally (F5) to produce a generally reliable confidence interval. The maximum
uncertainty regarding the number of produced photons within each energy bin of widths 0.5, 1
and 2 keV is less than 2% in our simulations, the magnitude of which depends on the number
of simulated electrons (in our case 4 × 107).
The x-ray beam is produced as electrons slow down in the anode and encounter some
attenuation in the anode material depending on the anode angle and the beam direction. The
x-rays propagate isotropically from the point of production, but the direction towards the Be
window is the only one of interest for measurement of the x-ray spectra. After passing through
the Be window, the x-ray spectrum passes through the additional filter material and air for
further attenuation of the x-ray beam before the measurement point. Table 1 summarizes
all the x-ray tube parameters investigated in the experiments carried out in this work. The
simulated x-ray spectra using MCNP4C were compared with experimental measurements and
spectra calculated by IPEM report number 78 described below. The comparative assessment
encompassed calculation of transmission curves based on both computational models. The
calculation of transmission curves involved running the code for different thicknesses of
aluminium filter incrementing in 1 mm steps in radiology and 0.1 mm in mammography.
To convert the photon spectra to kerma in air, the total number of transmitted photons in
each energy bin for each thickness was multiplied by a related photon to kerma conversion
factor (ICRU 1989) and then normalized relative to the air kerma without the filter being
present.
Quantitative evaluation of the differences between Monte Carlo simulated and calculated
spectra was performed using statistical analysis. The student’s t-test values and the
corresponding significance levels associated with the student’s analysis (two-tailed test)
Monte Carlo simulation of x-ray spectra using MCNP4C 4905

Table 1. Summary of x-ray tube parameters investigated in this work using MCNP4C-based Monte
Carlo simulations.
Figure Tube voltage (kV) Target/Angle Filter (mm) FSDa (mm)

2ab 80–140 W/12◦ 1 Be/2.5 Al 750

2bb 140 W/12◦ 1 Be/2.5 Al/0.1 Cu 750
3ab 50 W/12◦ 1.2 Al 3500
3bb 80 W/12◦ 1.2 Al 3500
3cb 100 W/12◦ 1.2 Al 3500
4b 80 (ripple 0–30%) W/12◦ 1.2 Al 750
5ab 30 Mo/10◦ 0.5 Be/0.03 Mo 500
5bb 30 Mo/10◦ 0.5 Be/0.025 Rh 500
5cb 30 W/14◦ 0.5 Be/1.2 Al 750
6a,bb 30 Mo/10◦ 0.5 Be/0.03 Mo 500
7ac 50–140 W/12◦ 1.2 Al/0–20 Al 750
7bc 30 Mo/12◦ 1 Be/0.035 Mo/0–2 Al 500
8a,b,cb 100 W/12◦ 1.2 Al 3500
9a,b,cb 100 W/6◦ –14◦ 1.2 Al 750
10ad 100 W/12◦ 1.2 Al 3500
10bd 100 W/12◦ 1.2 Al 200
10cd 100 W/8◦ –12◦ 1.2 Al 750
11ad 80–140 W/6◦ –18◦ 1 Be/2.5 Al 750
11bd 25–35 Mo/6◦ –18◦ 0.5 Be/0.03 Mo 1000
a Distance between focal spot and measurement point.
b X-ray spectrum data.
c Transmission curve data.
d Radiation intensity data.

were calculated for the MCNP4C simulated spectra as compared to the IPEM report and
experimental measurements. If the calculated t-value is greater than the critical t-value,
the null hypothesis of no statistically significant difference (at the 95% confidence level) is
rejected. It is worth pointing out that failure to prove statistically significant differences is
not sufficient to confirm that the results are statistically identical. Generally, the paired t-test
showed no statistically significant differences between the MCNP and IPEM spectra for all
experiments performed in this work.
One unfortunate consequence of the line-focus principle is that the radiation intensity
on the cathode side of the x-ray field is higher than that on the anode side. The intensity
distribution falling on a plane at right angles on the axis of the window (e.g., a film or a
screen) is not uniform owing to the anode heel effect (Bushong 1998). In our experiments,
the anode heel effect and off-axis x-ray spectra were assessed for different anode angles and
target materials and the results were compared with EGS4-based Monte Carlo simulations and
measured data published by Bhat et al (1999).

2.3. IPEM report no. 78

The original version of the catalogue based on a semi-empirical model for computing x-ray
spectra (Birch and Marshall 1979) was published in 1979 and provided essential data useful
for applications in diagnostic radiology and mammography (Birch et al 1979). The current
electronic version (Cranley et al 1997) contains sets of radiology and mammography x-ray
spectra with much wider ranges than the previous version. This version uses the XCOM photon
cross section library (Berger and Hubbell 1987) to calculate linear attenuation coefficients of
various materials. The unattenuated photon spectra are given for tungsten targets, tube potential
4906 M R Ay et al

1.20E-08

1.00E-08 XNUM=1
XNUM=0.01

Normalized No./keV
8.00E-09

6.00E-09

4.00E-09

2.00E-09

0.00E+00
0 5 10 15 20 25 30 35
(a) Energy (keV)

3.50E-01

MCNP
3.00E-01
IPEM

2.50E-01
Normalized No./keV

2.00E-01

1.50E-01

1.00E-01

5.00E-02

0.00E+00
0 5 10 15 20 25 30 35
(b) Energy (keV)

Figure 6. (a) Plots of 30 kV x-ray spectra simulated by MCNP4C for Mo target with 0.5 mm Be
and 0.03 mm Mo filter produced for XNUM = 1 and 0.01. (b) Comparison between x-ray spectra
produced by MCNP4C with XNUM = 1 and IPEM report 78 after manual setting of the intensity
of characteristic photons in the MCNP4C spectra to match the value computed by IPEM.

from 30 kV to 150 kV, and target angles from 6◦ to 22◦ . The ripple value can be changed from
0 to 30%. Constant potential mammographic spectra are provided from 25 kV to 32 kV for
molybdenum and rhodium targets for target angles ranging between 9◦ and 23◦ . All spectra
are provided at energy bin width of 0.5 keV (Cranley et al 1997). The IPEM report 78 was
used as reference to compare with the MCNP simulations because of its popularity and wide
availability (Ng et al 2000).

3. Results and discussion

3.1. X-ray spectra in diagnostic radiology and mammography

One of the most important parameters influencing the quality of the x-ray spectrum is
filtration. The produced x-ray beam after attenuation in the target passes through the
Monte Carlo simulation of x-ray spectra using MCNP4C 4907

1
MCNP
IPEM

Fraction of Transmitted kerma

140 kV
0.1 120 kV

100 kV

80 kV

0.01

50 kV

0.001
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

(a) Thickness of Aluminium (mm)

1
MCNP
IPEM
Fraction of Transmitted kerma

0.1

0.01
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1

(b) Thickness of Aluminium (mm)

Figure 7. Comparison of transmission curves produced by MCNP4C and IPEM report 78 for
(a) tungsten target between 50 and 140 kV and (b) molybdenum target at 30 kV.

tube’s inherent filtration (e.g., Be window) and other filter materials (e.g., aluminium for
the attenuation of soft x-rays). Figure 2(a) shows simulated x-ray spectra according to the
set-up summarized in table 1 for different tube voltages and their comparison with IPEM report
no. 78. The uncertainty achieved in figure 2(a) by simulating 4 × 107 electrons is ∼2% and this
value decreases to ∼1% in figure 2(b) by simulating 1 × 108 electrons. Figure 2(b) shows
the simulated x-ray spectrum and its comparison with the IPEM report no.78 for 140 kV tube
voltage using a different filter combination, namely 2.5 mm Al, 1 mm Be and 0.1 mm Cu
additional filter.
The x-ray spectrum shows significant tungsten K x-rays at 58, 59.5, 67.5 and 69 keV, the
small shift in characteristic x-ray energy being the result of binning the data into 0.5 keV energy
intervals. It can be seen that our results have good agreement with the IPEM report with small
differences visible in the intensity of characteristic x-rays and the low energy range, namely
between 20 and 40 keV. The intensity of Kα1 (59 keV) and Kα2 (58 keV) x-ray production by
4908 M R Ay et al

5.00E-02
MCNP
4.50E-02
Measured (Bhat et al 1999)

4.00E-02

Normalized No./keV
3.50E-02

3.00E-02

2.50E-02

2.00E-02

1.50E-02

1.00E-02

5.00E-03

0.00E+00
0 20 40 60 80 100
(a) Energy (keV)

5.00E-02
MCNP
4.50E-02
Measured (Bhat et al 1999)

4.00E-02

3.50E-02
Normalized No./keV

3.00E-02

2.50E-02

2.00E-02

1.50E-02

1.00E-02

5.00E-03

0.00E+00
0 20 40 60 80 100

(b) Energy (keV)

Figure 8. Comparison of simulated and measured on- and off-axis x-ray spectra for 100 kV tube
voltage. (a) Central axis, (b) 6◦ cathode side and (c) 6◦ anode side.

MCNP4C is slightly higher than IPEM in tube voltages lower than 100 kV and this behaviour
is reversed for tube voltages greater than 100 kV while the intensity of Kβ1 (67 keV) and
Kβ2 (69 keV) characteristic photon production by IPEM is higher than MCNP4C for all tube
voltages. This is consistent with the observations made by Verhaegen et al (1999) using the
previous version of the code (MCNP4B). Characteristic photons in MCNP are created by the
electron impact ionization (EII) process. However, the model overestimates the total number
of EII characteristic photons especially in the mammography energy range. This is regulated
by the parameter XNUM on the PHYS:E card, which is used to control the sampling of x-ray
photons produced along electron substeps.
Figure 3 compares simulated spectra (2 keV energy bin) with experimental spectra
published by Fewell et al (1981) and Bhat et al (1998) for different tube voltages. The
simulated spectra have higher intensity in characteristic x-rays for the reasons mentioned
Monte Carlo simulation of x-ray spectra using MCNP4C 4909

5.00E-02
MCNP

4.50E-02
Measured (Bhat et al 1999)

4.00E-02

3.50E-02

Normalized No./keV
3.00E-02

2.50E-02

2.00E-02

1.50E-02

1.00E-02

5.00E-03

0.00E+00
0 20 40 60 80 100

(c) Energy (keV)

Figure 8. (Continued.)

above. Note that a pure tungsten target was used in our simulation, while the targets used in
the experimental measurements have some rhenium and molybdenum impurity. It is worth
emphasizing that both el03 and el1 treatments only take into account the highest Z component
of the material for production of k-shell x-rays in MCNP (Briesmeister 2000). Thus, the
tungsten k-shell masks the x-ray characteristic of added low Z impurities. Figure 4 shows the
effect of tube voltage ripple on production of simulated x-ray spectra and its comparison with
IPEM report no. 78. Some authors modelled the tube voltage ripple by combining a number of
spectra generated at different constant tube potentials, each one being weighted according to
the time for which that tube potential occurs (Boone and Seibert 1997, Cranley et al 1997). In
our experiments, we modelled the tube voltage ripple during Monte Carlo simulation of x-ray
spectra by uniformly sampling from the tube voltage waveform (equation (2)). Similar to
the results reported above, there is good agreement between simulated and calculated spectra
apart from the small shift in characteristic photon energy for the reasons explained above. A
30% ripple is probably higher than most inverter generators as most of them produce ripple
factors between 5% and 15% (Boone and Seibert 1997).
Comparisons of the simulated x-ray spectra using MCNP4C with IPEM report 78 for
different target/filter material combinations including molybdenum and tungsten as targets
and molybdenum, rhodium and aluminium as filters are shown in figure 5. Although there are
no statistically significant differences between MCNP4C and IPEM data in Mo target material,
the low energy x-ray (<19.5 keV) intensity calculated by IPEM is higher than MCNP4C, which
is again the result of the significant overestimation of the intensity of characteristic x-rays in
MCNP4C following the normalization procedure. The same observations were reported by
Wilkinson et al (2001), where the characteristic x-rays in their measured spectrum at 20 keV
had a higher intensity in comparison with the IPEM model predictions.
Figure 5(b) shows that the rhodium filter transmits more bremsstrahlung radiation than the
Mo filter especially for energies greater than 20 keV, because the attenuation of the Rh filter in
this energy range is lower than the attenuation of the Mo filter. Figure 5(c) shows the tungsten
target spectra where the spectrum generated by MCNP4C has good agreement with IPEM
in the energy range <17 keV, but the intensity of the spectrum in the energy range 17–25 keV
is lower than IPEM. This behaviour is reversed for energies >25 keV.
4910 M R Ay et al

5.00E-09
6 Degree
8 Degree
4.50E-09
10 Degree
12 Degree
4.00E-09 14 Degree

3.50E-09
Normalized No./keV
3.00E-09

2.50E-09

2.00E-09

1.50E-09

1.00E-09

5.00E-10

0.00E+00
0 20 40 60 80 100
(a) Energy (keV)

5.00E-09
6 Degree
8 Degree
4.50E-09 10 Degree
12 Degree
14 Degree
4.00E-09

3.50E-09
Normalized No./keV

3.00E-09

2.50E-09

2.00E-09

1.50E-09

1.00E-09

5.00E-10

0.00E+00
0 20 40 60 80 100
(b) Energy (keV)

Figure 9. Simulated x-ray spectra for 140 kV tube voltage for various target angles. (a) Central
axis. (b) 6◦ cathode side. (c) 6◦ anode side.

The optimal adjustment of the XNUM parameters in the PHYS:E card to reduce the
intensity of characteristic x-ray production proved to be a difficult issue. Figure 6(a) shows
the difference between the simulated spectra with different values of XNUM (1 and 0.01)
with the set-up used in figure 5(a). It can be shown that the difference between MCNP
simulated spectra and IPEM spectra in the low energy range (figures 5(a) and (b)) is the result
of the overestimation of characteristic photons following the normalization procedure. Good
agreement between the spectra has been achieved (figure 6(b)) after manual adjustment of the
characteristic x-ray intensity in the MCNP spectrum (XNUM = 1) to match the value of the
IPEM spectra.
The difference in the quality of x-ray spectra produced by MCNP4C and IPEM report
no. 78 is further illustrated in figure 7 which shows the transmission curves through an
aluminium filter computed from various spectra in increments of 1 mm in diagnostic radiology
Monte Carlo simulation of x-ray spectra using MCNP4C 4911

5.00E-09
6 Degree
8 Degree
4.50E-09 10 Degree
12 Degree
14 Degree
4.00E-09

3.50E-09
Normalized No./keV
3.00E-09

2.50E-09

2.00E-09

1.50E-09

1.00E-09

5.00E-10

0.00E+00
0 20 40 60 80 100
(c) Energy (keV)

Figure 9. (Continued.)

Table 2. Comparison between measured, calculated (IPEM report 78) and simulated (MCNP4C)
HVLs and mean spectrum energy for spectra produced using 12◦ target angle and different tube
voltages. The percentage difference between IPEM and MCNP4C estimates is also shown.

Mean spectrum
energy (keV) First HVL (mm Al)
Target Filter
kV material (mm) IPEM MCNP Diff (%) Bhat (1998) Fewell (1981) IPEM MCNP Diff (%)
30 Mo 0.5 Be 16.8 17.0 −1.2 – – 0.32 0.35 −9.4
0.035 Mo
50 W 1.2 Al 29.9 29.4 +1.7 1.32 – 1.19 1.18 +0.8
80 W 1.2 Al 40.5 40.4 +0.3 1.80 1.81 1.97 1.95 +1.0
100 W 1.2 Al 46.9 46.2 +1.5 2.35 2.29 2.64 2.51 +4.9
120 W 1.2 Al 52.6 51.6 +1.9 – – 3.38 3.17 +6.2
140 W 1.2 Al 57.4 56.3 +1.9 – – 4.19 3.75 +10.5

and 0.1 mm in mammography. The transmission curves produced by MCNP4C have good
agreement with the IPEM report especially for tube voltages of 50 kV and 80 kV. The
systematic discrepancy for higher tube voltages is the result of systematic differences between
the corresponding spectra (figure 2). The average and maximum differences in transmission
curves for energies between 50 and 140 kV (five energies) are −2.7, 3.2, 6, 6.7, 7.31% and
−5.9, 3.8, 7.5, 8.4, 9.2%, respectively (figure 7(a)).
Table 2 compares the mean spectrum energy and HVL for different tube voltages as
measured by physical experiments and calculated using IPEM and MCNP4C spectra. The
percentage differences vary between −1.2% to 1.9% and −9.4% to 10.5% for mean spectrum
energy and HVLs, respectively, for all tube voltages and targets/filters investigated. The results
indicate that the percentage difference between HVLs calculated from MCNP4C spectra and
IPEM spectra increases with increasing tube voltage. Moreover, the mean energy of IPEM
spectra in the diagnostic radiology energy range is higher than spectra calculated by MCNP4C.
The amplitude of this difference increases with increasing tube voltage, thus increasing the
differences between transmission curves. The higher quality of IPEM spectra for tube voltages
4912 M R Ay et al

1.20E+00
MCNP
EGS4 (Bhat et al 1999)
Measured (Bhat et al 1999)

1.00E+00

Relative Exposure
8.00E-01

6.00E-01

4.00E-01

2.00E-01

(Anode Side) (Cathode Side)

0.00E+00
-15 -10 -5 0 5 10 15
(a) Angle from Central Axis (degree)

1.20E+00

1.00E+00

8.00E-01
Relative Exposure

6.00E-01

4.00E-01

2.00E-01

(Anode Side) (Cathode Side)

0.00E+00
-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90

(b) Angle from Central Axis (degree)

Figure 10. (a) Comparison of simulated and measured relative exposure values in different
directions relative to the central axis (calculation points B in figure 1). (b) Relative exposure around
the target (calculation points A in figure 1). (c) Normalized exposure on the axis perpendicular to
the anode–cathode axis.

higher than 80 kV induces higher values in transmission curves in comparison with MCNP4C
spectra.
In our simulation of x-ray spectra resulting from the use of different target/filter
combinations in mammography, the tube voltage was assumed to be 30 kV. The transmission
curve for the spectrum produced from a molybdenum target with 1 mm Be and 0.035 mm
Mo additional filter was calculated. An uncertainty of 0.6% was reached by simulating 3 ×
107 electrons. In the mammography energy range, the quality of MCNP4C spectra is higher
than IPEM, thus the IPEM transmission curves have lower values compared to those estimated
by MCNP4C. The average and maximum differences observed are −7.6% and −9.8% at
tube voltage 30 kV in mammography (figure 7(b)). To further investigate the validity of our
simulation results, we compared our calculated HVL with measured data published by the
Monte Carlo simulation of x-ray spectra using MCNP4C 4913

5.40E-09

8 Degree
5.30E-09 12 Degree

Normalized Exposure (MeV/cm )

2
5.20E-09

5.10E-09

5.00E-09

4.90E-09

4.80E-09

4.70E-09

4.60E-09

4.50E-09

4.40E-09
-20 -15 -10 -5 0 5 10 15 20
(c) Angle from Central Axis (degree)

Figure 10. (Continued.)

SSDL laboratory of the IAEA (Pernieka et al 1997). They reported 0.348 mm Al for first
HVL while we calculated 0.35 mm (−0.6% difference). Note that there are some differences
between their experimental set-up (0.8 mm Be filter, 0.03 mm Mo filter, 20◦ Mo target angle)
and our simulation set-up (1 mm Be filter, 0.035 mm Mo filter, 12◦ Mo target angle). According
to our estimation, the 8◦ difference in target angle produces 3.8% difference in the first
HVL.

3.2. Assessment of anode heel effect in diagnostic radiology and mammography

Figure 8 shows the on- and off-axis spectra for a tungsten target at 6◦ anode side and 6◦
cathode side and compares them with experimental measurements (Bhat et al 1999). The
shape of the anode side spectrum after passing through the target skewed towards the higher
energy with some distortion in the characteristic tungsten x-ray at 69.5 keV (figure 8(c)).
Figure 9 shows the central axis and off-axis x-ray spectra for different target angles. The
anode side spectrum encounters more attenuation when using small target angles while the
cathode side spectra are approximately similar for all target angles except for a slightly higher
attenuation for small target angles in the low energy range.
Figure 10(a) shows the variation of relative exposure with respect to the central axis
(calculation points B in figure 1) as simulated by the MCNP4C code and compares it with
experimental measurements and EGS4-based Monte Carlo simulations published by Bhat
et al (1999). The average difference between EGS4 and MCNP simulations for 18 different
angles is 0.8%. The maximum difference is −5.5% at 6◦ anode side, while the maximum
difference between MCNP and measured data is −8.2% in the same position. Generally
there is good agreement between the relative exposure estimated by MCNP, EGS4 and the
experimental data on the cathode side. A small discrepancy (−1.6%) is, however, visible
starting from 4◦ anode side due in our opinion to differences in target dimensions used in both
experiments. The radiation escaping from the x-ray tube housing through areas other than the
desired x-ray window increases the dose to the patient and surroundings. Figure 10(b) shows
the relative exposure around the target at distance 200 mm from the focal spot, which is useful
for calculation of x-ray tube shielding. The exposure has been calculated after attenuation of
produced x-ray spectra by a 1.2 mm Al filter (calculation points A in figure 1). Obviously, the
4914 M R Ay et al

2.50E-07
6 Degree
10 Degree
14 Degree
18 Degree
2.00E-07

Normalized Exposure (MeV/cm )

2
140 KV
1.50E-07

120 KV

1.00E-07

5.00E-08
80 KV

0.00E+00
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
(a) Angle from Central Axis (degree)

1.20E-08
6 Degree
10 Degree
14 Degree
18 Degree
1.00E-08
Normalized Exposure (MeV/cm )
2

8.00E-09 35 KV

6.00E-09
30 KV

4.00E-09

25 KV
2.00E-09

0.00E+00
-12 -7 -2 3 8 13 18 23
(b) Angle from Central Axis (degree)

Figure 11. Illustration of anode heel effect for different anode angles for (a) tungsten target
with tube voltages between 80 and 140 kV and for (b) molybdenum target with the tube voltages
between 25 and 35 kV.

relative exposure on the cathode side is higher than that on the anode side owing to attenuation
of the x-ray spectra in the target.
Another problem in x-ray imaging is nonuniformity of exposure in the direction
perpendicular to the anode–cathode axis, bearing in mind that the field of view is usually
rectangular. Figure 10(c) shows the variation of exposure (normalized to 2 × 107 simulated
electrons) on this axis for 8◦ and 12◦ target angles at 100 kV tube voltage. The off-axis
exposure is identical on both sides of the central axis and the absorption at small target angles
is higher than that at large target angles. In addition, we have calculated the off-axis spectra in
both cases. The results show that the off-axis and central axis spectra have approximately the
same shape without any distortion similar to data shown in figure 9. Our further investigation
Monte Carlo simulation of x-ray spectra using MCNP4C 4915

Table 3. Variation in radiation exposure for various target angles and tube voltages in the useful
x-ray beam.

Tube Target Difference in radiation exposure

voltage (kV) material in the useful beama (%)
Target angle
6◦ 10◦ 14◦ 18◦
25 Mo 96.0 11.7 3.7 1.3
30 Mo 96.7 15.7 5.7 2.7
35 Mo 97.4 19.8 8.0 4.6
80 W 92.0 21.6 9.5 4.4
120 W 94.2 30.7 14.3 8.4
140 W 94.4 33.2 17.5 11.15
a From 6◦ anode side until 6◦ cathode side.

of the influence of focal spot size on anode heel effect performed by calculating the heel effect
for different focal spot sizes (0.3, 0.6, 1.2 mm) indicated that its effect is negligible.
Figure 11(a) shows the anode heel effect for different target angles and tube voltages.
◦
The relative difference in radiation exposure in the useful x-ray beam (6 anode side until 6◦
cathode side) calculated from figure 11(a) is shown in table 3. To demonstrate the validity of
results presented in this table, we have compared our simulated data for 12◦ target angle and
100 kV with experimentally measured and EGS4 Monte Carlo simulation results published by
Bhat et al (1999). It turns out that the difference in radiation exposure calculated by MCNP4C
is 20.5% while this value is 20.0% for EGS4 and 22.0% for measured data (figure 10(a)).
In theory, the anode heel effect is of considerable importance in mammography. It is
expected that the conic shape of breasts would require greater radiation intensity near the
chest wall rather than to the nipple side so that near uniform exposure of the image receptor
will occur. This could be accomplished by positioning the cathode on the chest wall side.
However, in practice this is not necessary because compression of the breast ensures that a
uniform thickness of tissue is imaged (Bushong 1998). Figure 11(b) shows the anode heel
effect for different target angles and tube voltages for Mo/Mo target/filter combinations. The
difference in radiation exposure in the useful x-ray beam calculated from figure 11(b) is shown
in table 3. It can be seen that the radiation exposure difference in the mammography energy
range increases by increasing the tube voltage and decreasing the target angle.

4. Conclusion

Monte Carlo simulation of x-ray spectra has been used extensively in different medical
imaging applications including assessment of image quality, optimization of system design
and absorbed dose calculation (Caon et al 1998, Dance et al 2000, Spyrou et al 2002,
Zaidi and Sgouros 2002). This study used the MCNP4C general-purpose Monte Carlo
code for generating x-ray spectra in diagnostic radiology and mammography. Although the
simulation of x-ray spectra using the Monte Carlo method is time consuming, the generated
x-ray spectra provide detailed information about particles’ interaction with different target
and filter combinations. This information is useful for x-ray tube design and development
of new target/filter combinations to improve image quality in diagnostic radiology and
mammography.
The generated x-ray spectra in this work have been verified against IPEM report number
78, experimental measurements and EGS4 Monte Carlo simulations for validation. Although
4916 M R Ay et al

the paired t-test results show no statistically significant differences between simulated and
reference spectra, the intensity of characteristic x-rays in MCNP4C simulated spectra is
slightly higher than that in IPEM spectra for tube voltages <100 kV, and this behaviour
is reversed for tube voltages >100 kV in the diagnostic radiology energy range. Whereas
MCNP produces a significantly higher intensity of characteristic x-rays compared to IPEM
in mammography when using a Mo target. This discrepancy can be tackled by appropriate
adjustment of the XNUM parameter. The results indicate that the MCNP4C general purpose
Monte Carlo code with some small adjustment in the appropriate MCNP cards is a useful tool
for generating diagnostic radiology and mammography x-ray spectra and investigation of the
heel effect.

Acknowledgments

This work was supported by the AmirKabir University of Technology, Department of Physics
and Nuclear Sciences and the Swiss National Science Foundation under grant SNSF 3152A0-
102143. The authors are indebted to M Bhat for providing measured and EGS4 simulated
data.

References

Acosta E, Liovet X, Coleoni E, Riveros J A and Salvat F 1998 Monte Carlo simulation of x-ray emission by kilovolt
electron bombardment J. Appl. Phys. 83 6038–49
Antonuk L E et al 1997 Empirical investigation of the signal performance of a high-resolution, indirect detection,
active matrix flat-panel imager (AMFPI) for fluoroscopic and radiographic operation Med. Phys. 24 51–70
Ay M, Sarkar S, Shahriari M, Sardari D and Zaidi H 2004 Comparative assessment of different computational
models for generation of x-ray spectra in diagnostic radiology and mammography Proc. IEEE Nuclear Science
Symposium and Medical Imaging Conference (Rome, Oct 19–22) at press
Ben Omrane L, Verhaegen F, Chahed N and Mtimet S 2003 An investigation of entrance surface dose calculations for
diagnostic radiology using Monte Carlo simulations and radiotherapy dosimetry formalisms Phys. Med. Biol.
48 1809–24
Berger M J and Hubbell J H 1987 XCOM: photon cross sections on a personal computer NBSIR 87-3597 (National
Bureau of Standards (US))
Bhat M, Pattison J, Bibbo G and Caon M 1998 Diagnostic x-ray spectra: a comparison of spectra generated by
different computational methods with a measured spectrum Med. Phys. 25 114–20
Bhat M, Pattison J, Bibbo G and Caon M 1999 Off-axis x-ray spectra: a comparison of Monte Carlo simulated and
computed x-ray spectra with measured spectra Med. Phys. 26 303–9
Birch R and Marshall M 1979 Computation of bremsstrahlung x-ray spectra and comparison with spectra measured
with a Ge(Li) detector Phys. Med. Biol. 24 505–17
Birch R, Marshall M and Ardran G M 1979 Catalogue of spectral data for diagnostic x-rays hospital physicists
Association Scientific Report Series 30
Blough M M, Waggener R G, Payne W H and Terry J A 1998 Calculated mammographic spectra confirmed with
attenuation curves for molybdenum, rhodium, and tungsten targets Med. Phys. 25 1605–12
Boone J M 1988 The three parameter equivalent spectra as an index of beam quality Med. Phys. 15 304–10
Boone J M, Fewell T R and Jennings R J 1997 Molybdenum, rhodium, and tungsten anode spectral models using
interpolating polynomials with application to mammography Med. Phys. 24 1863–74
Boone J M and Seibert J A 1997 An accurate method for computer-generating tungsten anode x-ray spectra from 30
to 140 kV Med. Phys. 24 1661–70
Briesmeister J F 2000 MCNP—a general Monte Carlo N-particle transport code. version 4C Los Alamos National
Laboratory Report LA-13709-M (Los Alamos, NM)
Bushong S C 1998 Radiologic Science for Technologists: Physics, Biology and Protection (St. Louis: Mosby Co)
Caon M, Bibbo G, Pattison J and Bhat M 1998 The effect on dose to computed tomography phantoms of varying the
theoretical x-ray spectrum: a comparison of four diagnostic x-ray spectrum calculating codes Med. Phys. 25
1021–7
Monte Carlo simulation of x-ray spectra using MCNP4C 4917

Cranley K, Gilmore B J, Fogarty G W A and Desponds L 1997 IPEM Report 78: Catalogue of Diagnostic X-ray
Spectra and Other Data (CD-Rom Edition 1997) (Electronic Version prepared by D Sutton) (York: The Institute
of Physics and Engineering in Medicine (IPEM))
Dance D et al 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 Br. J. Radiol. 73 1056–67
Fewell T R and Shuping R E 1977 Photon energy distribution of some typical diagnostic x-ray beams Med. Phys. 4
187–97
Fewell T R and Shuping R E 1978 Handbook of Mammography Spectra (HEW Publication) (FDA) 79-8071
Fewell T R, Shuping R E and Healy K E 1981 Handbook of Computed Tomography X-Ray Spectra (HHS Publication)
(FDA) 81-8162 (Washington, DC: US Govt. Printing Office)
Hughes H G 1997 Status of electron transport in MCNP Los Alamos National Laboratory LA-UR-97-1368 (Los
Alamos, NM)
ICRU 1989 Tissue Substitutes in Radiation Dosimetry and Measurement ICRU Report 44 (Bethesda, MD: International
Commission on Radiological Units and Measurements (ICRU))
Kramers H A 1923 On the theory of x-ray absorption and of the continuous x-ray spectrum Phil. Mag. 46 836–71
Kulkarni R N and Supe S J 1984 Monte Carlo calculations of mammographic x-ray spectra Phys. Med. Biol. 29
185–90
Laitano R F, Pani R and Pellegrini R 1991 Determination of x-ray spectra and of the scattered component up to
300 kV Med. Phys. 18 934–8
Mercier J R et al 2000 Modification and benchmarking of MCNP for low-energy tungsten spectra Med. Phys. 27
2680–7
Ng K P, Kwok C S and Tang F H 2000 Monte Carlo simulation of x-ray spectra in mammography Phys. Med. Biol.
45 1309–18
O’Meara J M, Chettle D R, McNeill F E, Prestwich W V and Svensson C E 1998 Monte Carlo simulation of
source-excited in vivo x-ray fluorescence measurements of heavy metals Phys. Med. Biol. 43 1413–28
Pernieka F, Andreo P, Meghziefene A, Czap L and Girzikowsky R 1997 Standards for radiation protection and
diagnostic radiology at the IAEA dosimetry laboratory SSDL Newsletter 17 (Vienna: International Atomic
Energy Agency)
Spyrou G, Tzanakos G, Nikiforides G and Panayiotakis G 2002 A Monte Carlo simulation model of mammographic
imaging with x-ray sources of finite dimensions Phys. Med. Biol. 47 917–33
Tucker D M, Barnes G T and Chakraborty D P 1991 Semiempirical model for generating tungsten target x-ray spectra
Med. Phys. 18 211–8
Verhaegen F and Castellano I A 2002 Microdosimetric characterisation of 28 kVp Mo/Mo, Rh/Rh, Rh/Al, W/Rh
and Mo/Rh mammography x ray spectra Radiat. Prot. Dosim. 99 393–6
Verhaegen F, Nahum A E, Van de Putte S and Namito Y 1999 Monte Carlo modelling of radiotherapy kV x-ray units
Phys. Med. Biol. 44 1767–89
Wilkinson L E, Johnston P N and Heggie J C 2001 A comparison of mammography spectral measurements with
spectra produced using several different mathematical models Phys. Med. Biol. 46 1575–89
Zaidi H and Sgouros G (ed) 2002 Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine (Bristol:
Institute of Physics Publishing)