Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

A Monte Carlo dose calculation algorithm for proton therapy

Matthias Fippela)
Abteilung für Medizinische Physik, Universitätsklinikum Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen,
Germany
Martin Soukup
Department of Dosimetry and Application of Ionizing Radiation, Czech Technical University in Prague,
Brehova 7, 11519 Praha, Czech Republic
共Received 17 March 2004; revised 13 May 2004; accepted for publication 18 May 2004;
published 23 July 2004兲
A Monte Carlo 共MC兲 code 共VMCpro兲 for treatment planning in proton beam therapy of cancer is
introduced. It is based on ideas of the Voxel Monte Carlo algorithm for photons and electrons and
is applicable to human tissue for clinical proton energies. In the present paper the implementation
of electromagnetic and nuclear interactions is described. They are modeled by a Class II condensed
history algorithm with continuous energy loss, ionization, multiple scattering, range straggling,
␦-electron transport, nuclear elastic proton nucleus scattering and inelastic proton nucleus reactions.
VMCpro is faster than the general purpose MC codes FLUKA by a factor of 13 and GEANT4 by a
factor of 35 for simulations in a phantom with inhomogeneities. For dose calculations in patients
the speed improvement is larger, because VMCpro has only a weak dependency on the heteroge-
neity of the calculation grid. Dose distributions produced with VMCpro are in agreement with
GEANT4 results. Integrated or broad beam depth dose curves show maximum deviations not larger
than 1% or 0.5 mm in regions with large dose gradients for the examples presented here. © 2004
American Association of Physicists in Medicine. 关DOI: 10.1118/1.1769631兴

Key words: proton therapy, Monte Carlo, treatment planning

I. INTRODUCTION rithms, because ray-tracing must be performed for time de-


pendent density distributions. This increases calculation time
Presently, pencil beam algorithms1– 8 are employed for dose and may be clinically impractical. For MC algorithms, on the
calculation in proton therapy. They have the advantage of other hand, it is easy to model these effects by choosing the
short calculation times, which is especially useful for inten- density in each voxel randomly from a pre-defined uncer-
sity modulated proton therapy.9 On the other hand, pencil tainty distribution before starting the particle history. This
beam algorithms show accuracy limitations mainly because can be realized without increasing the calculation time com-
of the one-dimensional density scaling of proton pencil pared to static Monte Carlo. Of course, the distribution func-
beams in water.10 To overcome these difficulties more so- tion of densities in each voxel must be known, e.g., by a time
phisticated scaling methods have been introduced.10,11 These dependent set of computerized tomography 共CT兲 cubes.
methods could improve dose calculation considerably. How- Therefore, in near future MC algorithms will be preferred for
ever, close to material interfaces there are remaining discrep- treatment planning in proton therapy if they are fast enough.
ancies to Monte Carlo calculated dose distributions.11 The Recently, Kohno et al. developed a simplified Monte
reason is, multiple elastic Coulomb scattering in heteroge- Carlo 共SMC兲 planning algorithm for proton beams based on
neous media as well as elastic and inelastic nuclear reactions measured depth dose distributions in water.12 They stated,
can be implemented in pencil beam algorithms only approxi- that in relatively short calculation time SMC agrees well
mately. Further improvement of proton dose calculation can with experimental results in a heterogeneous phantom. How-
be achieved using Monte Carlo 共MC兲 algorithms. ever, further investigations will be necessary to quantify the
Some of the most serious problems in proton therapy to- calculation time as well as the influence of these simplifica-
day are patient set-up errors and organ motion. These uncer- tions on the results.
tainties have a large impact on the treatment success, because Another option is the implementation of established gen-
the dose of proton beams can be placed with high precision eral purpose MC codes like GEANT4,13 FLUKA,14,15 or MCNP16
to specific locations within the patient. A small displacement for treatment planning. However, as we will show in Sec.
of the Bragg–Peak can easily lead to tumor underdosage as III E, these codes are still too slow for this type of simula-
well as overdosage to organs at risk. Presently these effects tions. For proton therapy treatment planning in future fast
are taken into account during treatment planning by adding and accurate MC dose calculation algorithms are required.
margins to the clinical target volume. But in this manner the In the present paper a MC algorithm for proton beams,
high precision of proton dose distributions decreases. Thus, it called VMCpro, is introduced. Here the implementation of
will be extremely useful to reduce the margins by taking electromagnetic and nuclear interactions is described. In the
motion effects into account during treatment planning. How- next section the properties of the transport model are ex-
ever, they are difficult to simulate with pencil beam algo- plained. It is discussed how the cross sections and transport

2263 Med. Phys. 31 „8…, August 2004 0094-2405Õ2004Õ31„8…Õ2263Õ11Õ$22.00 © 2004 Am. Assoc. Phys. Med. 2263
2264 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2264

parameters are determined from a given distribution of mass determined in energy space using the modified fictitious in-
densities. In Sec. III the new code is benchmarked using teraction method as described by Kawrakow21 taking into
GEANT4 and FLUKA in terms of accuracy and calculation account ionization, elastic and inelastic nuclear reactions.
time. Using the mass density ␳ in the present voxel, the stopping
power ratio to water f S ( ␳ ,T p ) can be determined 关see Sec.
II C, Eq. 共13兲兴. This allows scaling of the geometrical step
II. METHODS
length ⌬z to the corresponding step length in water ⌬z w
For treatment planning and optimization the models of according to:
patients are given by sets of computerized tomography 共CT兲
images. Using standard CT calibration techniques these sets ␳
⌬z w ⫽ f S 共 ␳ ,T p 兲 ⌬z. 共1兲
of images can be converted into three-dimensional 共3D兲 ␳w
grids of mass densities, called density matrices. Thus, the
The energy loss ⌬E is identical for protons with geometrical
transport of protons and secondary particles has to be simu-
step length ⌬z in a medium with density ␳ and protons of the
lated within voxelized geometries. This is identical to the
same energy with step length ⌬z w in water. Using the re-
situation in traditional radiation therapy where the transport
of photons and electrons must be simulated. This allows us e ) 共see Sec. II D兲
stricted stopping power in water L w (T p ,T min
and the integral
to reuse ideas developed for the Voxel Monte Carlo 共VMC兲
algorithm.17–19 Therefore, the Monte Carlo code for proton
beams, presented here, is called VMCpro. ⌬z w ⫽⫺ 冕 Tp
end
Tp dT ⬘p
L w 共 T ⬘p ,T min
e 兲
, 共2兲

A. Proton transport algorithm


it is possible to calculate the mean kinetic energy of the
The most important component of VMCpro is the algo- proton at the end of the step:
rithm to model the transport of protons with different ener-
gies through human tissue. This transport is influenced by p ⫽T p ⫺⌬E.
T end 共3兲
elastic and inelastic Coulomb collisions with atomic nuclei
and atomic electrons as well as elastic and inelastic nuclear The real proton energy T end
p at the end of the step fluctuates

reactions. around its mean value. In Sec. II E the implementation of this


Protons undergo a huge number 共more than 106 per cm) energy straggling is discussed.
of elastic Coulomb interactions. Simulating this collision by After moving the proton to the new position a multiple
collision 共analog Monte Carlo兲 would require too much cal- scattering angle is sampled 共see Sec. II F兲 and the proton is
culation time. Therefore, a Class II condensed history rotated according to this angle. The energy loss ⌬E is scored
algorithm20 is implemented with production of ␦-electrons in the present voxel and the new energy T end p is assigned to

共ionizations兲 above the threshold energy T min and continuous the proton. If there is a discrete interaction at the end of the
e
energy loss below T min min step this interaction is simulated 共see Secs. II B and II H兲.
e . The parameter T e can be specified
by the user. It influences the ionization cross section 共see This can lead to secondary particles, e.g., ␦-electrons or sec-
Sec. II B兲 and thus the number of ␦-electrons during the ondary protons. In this case the transport of these particles is
simulation. also simulated 共see, e.g., Sec. II G兲.
The transport simulation starts with a proton of definite The proton history ends if it leaves the calculation grid or
kinetic energy T p and momentum at the surface of the cal- its energy becomes smaller than T min p . Otherwise the proce-

culation grid. The kinetic energy of this proton is absorbed dure continues with the simulation of a new step.
locally if it is smaller than the minimum energy T min p . This
parameter can be defined by the user. We usually choose
p ⫽0.5 MeV. The range of protons with smaller energy
T min B. Ionization
共less than 0.01 mm in water兲 is negligible with respect to the
For kinematic reasons the maximum energy T max e transfer-
voxel resolution. Even in air cavities this choice will be suf-
able to free electrons during ionization interactions is given
ficient because of the small stopping power and nuclear in-
by 共see, e.g., ‘‘The Review of Particle Physics’’ 22兲:
teraction cross sections. T min
p is also used to control the end
of the simulation of each proton. The transport stops if the 2m e ␤ 2 ␥ 2
kinetic energy becomes smaller than the minimum energy. e ⫽
T max . 共4兲
1⫹2 ␥ m e /m p ⫹ 共 m e /m p 兲 2
There is also a maximum energy T max p ⫽500 MeV. It is used
as upper limit of proton transport data tabulations. Thus the Here, m e and m p are the electron and proton rest masses in
typical clinical energies 共from 60 to 270 MeV兲 are included. units of energy, respectively, ␤ is the ratio of proton velocity
The VMCpro transport procedure traces protons through to the velocity of light and the relativistic parameter ␥ is
the grid step by step. One step is defined by the distance given by ␥ ⫽E p /m p with the total energy of the proton E p
between two voxel boundaries if there is no discrete interac- ⫽T p ⫹m p .
tion in the present voxel. If there is a discrete interaction The differential macroscopic cross section for the produc-
共ionization or nuclear reaction兲 one step is until this interac- tion of ␦-electrons with kinetic energy T e ⬎T min
e is calculated
tion point. The distance to the discrete interaction point is by 共see, e.g., GEANT4, Physics Reference Manual 200323兲:

Medical Physics, Vol. 31, No. 8, August 2004


2265 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2265

d⌺
dT e
ne 1 Te

⫽2 ␲ r 2e m e 2 2 1⫺ ␤ 2 max ⫹
␤ Te Te
T 2e
2E 2p
, 册 共5兲

for materials with electron density n e , r e is the electron ra-


e ⭐Te . Using Eq. 共5兲
dius. This cross section is zero for T max min

the total macroscopic cross section

e 兲⫽
⌺ 共 n e ,T p ,T min 冕 Te
min
Te
max
dT e
d⌺
dT e
, 共6兲

and the first moment

e 兲⫽
M 共 n e ,T p ,T min 冕 max
Te
min
Te
dT e T e
d⌺
dT e
, 共7兲

can be calculated analytically. Equation 共5兲 also provides a


rejection method to sample the kinetic energy of the
␦-electron. First of all, this energy is sampled from distribu-
tion
FIG. 1. Mass stopping power ratios of all materials from ICRU Report 46
T min
e Te
max
1 共Ref. 24兲 to the mass stopping power of water for 10 MeV and 100 MeV
f 共 Te兲⫽ , 共8兲 protons 共crosses兲. The lines represent the fit of Eq. 共13兲 for both energies.
e ⫺T e
T max min 2
Te
i.e., with ␩ being a uniformly distributed random number in
关0:1兴, T e is calculated according to: ␳ w S 共 ␳ ,T p 兲
f S 共 ␳ ,T p 兲 ⫽ . 共12兲
T min
e Te
max
␳ S w共 T p 兲
T e⫽ . 共9兲
共 1⫺ ␩ 兲 T e ⫹ ␩ T min
max
e After analyzing the tabulations and calculations of ICRU Re-
Then, function port 46,24 ICRU Report 49,25 and the computer tool PSTAR26
we develop the following fit formula:
Te T 2e
1.0123⫺3.386 10⫺5 T p ⫹0.291共 1⫹T ⫺0.3421


g 共 T e 兲 ⫽1⫺ ␤ 2
⫹ , 共10兲 p 兲
T max 2E 2p ⫺0.7
e ⫻共 ␳ ⫺1 兲 for ␳ ⭓0.9
can be used as rejection weight. f S 共 ␳ ,T p 兲 ⫽ 0.9925 for ␳ ⫽0.26 共 lung兲
The polar scattering angle of the ␦-electron follows from 0.8815 for ␳ ⫽0.0012 共 air兲
the kinematics of this process. The azimuthal scattering interpolate for all other ␳ ⭐0.9,
angle is sampled from a uniform distribution. Because of the 共13兲
large proton mass compared to the electron mass a direction
with the kinetic proton energy T p in MeV and mass density ␳
change of the proton can be neglected. The proton direction
in g cm⫺3 . Figure 1 shows this function for two proton en-
mainly changes due to multiple elastic scatter with the
ergies 共10 and 100 MeV兲 compared to the ICRU Report 46
atomic nuclei 共see Sec. II F兲 or due to elastic nuclear colli-
data.24 Beside the outliers 共Gallstone, Protein, Carbohydrate
sions 共see Sec. II H兲.
and Urinary Stone兲, Eq. 共13兲 provides an accuracy of better
than 1%. This fit allows the scaling of Eq. 共1兲 during MC
C. Stopping power ratio
simulations based on mass densities only. Knowledge of the
The total proton stopping power is given by the sum of atomic composition in each voxel is not necessary. Equation
the electronic stopping power and the nuclear stopping 共13兲 can also be applied to pencil beam algorithms.
power. Above 1 MeV kinetic energy the nuclear stopping
power is negligible, i.e., protons are slowed down mainly
because of inelastic interactions with the atomic electrons. D. Restricted stopping power in water
The mass stopping power of a proton with kinetic energy The restricted stopping power in water L w (T p ,T min
e ) must
T p is defined by the energy loss dT p along the path length dz be known to calculate the energy loss of protons based on
and normalized by the mass density ␳, i.e., Eq. 共2兲. This value depends on the total stopping power in
S 共 ␳ ,T p 兲 1 dT p 共 ␳ ,T p 兲 water and the first moment of the ionization cross section
⫽ , 共11兲 due to:
␳ ␳ dz
e 兲 ⫽S w 共 T p 兲 ⫺M w 共 T p ,T e 兲 .
L w 共 T p ,T min 共14兲
min
with S( ␳ ,T p ) being the linear stopping power. For human
tissue, it is possible to express the stopping power of a ma- M w (T p ,T min
e ) is determined using Eq. 共7兲, therefore, the re-
terial with mass density ␳ as a function of the mass stopping stricted stopping power depends on T min
e . It models ioniza-
power in water S w (T p )/ ␳ w . For this reason we define the tion processes below the energy threshold. The total stopping
stopping power ratio power S w (T p ) is taken from PSTAR.26 L w (T p ,T min
e ) is cal-

Medical Physics, Vol. 31, No. 8, August 2004


2266 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2266

culated and tabulated during the initialization of VMCpro for


the given T min
e . In the neighborhood of T p it can be param-
eterized as
⫺[b(T p )⫺1]
L w 共 T p 兲 ⫽a 共 T p 兲 T p , 共15兲
with b(T p ) being an almost constant function. If the param-
eter b(T p ) is also calculated and tabulated prior to the MC
simulation, the integral of Eq. 共2兲 can be evaluated efficiently
and accurately.

E. Energy straggling
The production of secondary electrons during ionization
interactions leads to energy fluctuations of the primary pro-
tons. This influences the steepness of the depth dose distri-
bution behind the Bragg–Peak and must be taken into ac-
count during dose calculation. It is taken into account
automaticly for ␦-electron production above T mine . That is, if FIG. 2. Central axis depth dose distributions in water for monoenergetic
we choose T min
e small enough the dose is correctly predicted proton pencil beams of 100 MeV, calculated with GEANT4 version 5.1,
without the need for applying energy straggling. However, GEANT4 version 5.2 and FLUKA2002. Nuclear reactions are switched off.

the calculation time increases because a huge number of


␦-electrons is created. Even if the energy of these electrons is
absorbed locally, the computation time remains unacceptably sion 5.2 was released, we modified the value to E s
long. Therefore, we have implemented the sub-threshold ⫽12.0 MeV. This was necessary because the implementa-
straggling distribution introduced by Bohr27 and described in tion of multiple scattering in GEANT4 has changed between
ICRU Report 4925 and the GEANT4 Physics Reference the versions. By using FLUKA2002.414,15 instead of GEANT4 for
Manual.23 That is, we model the straggling function by a fitting, the parameter must be changed to E s ⫽12.9 MeV.
Gaussian with mean ⌬E from Eq. 共3兲 and variance: Figure 2 shows central axis depth dose curves in water of

冉 冊
100 MeV pencil beams with a voxel resolution of 1 mm
min共 T min max
e ,T e 兲 ␤2 calculated with GEANT4 version 5.1, GEANT4 version 5.2, and
⍀ 2 ⫽2 ␲ r 2e m e n e ⌬z 1⫺ . 共16兲
␤2 2 FLUKA2002.4, without nuclear reactions. The differences can
be explained by different multiple scattering distributions
The last term (1⫺ ␤ 2 /2) provides a relativistic correction to
implemented in GEANT4 and FLUKA. Strictly speaking, the
Bohr’s variance.
parameter E s should be determined by measurement or a
Using this approach we have calculated proton beam dose
theoretical computation. Therefore, additional investigations
distributions in water for various threshold energies T mine . will be necessary. But this is not the task of the present
The final result 共especially in the Bragg–Peak region兲 does
paper, we just provide a method to adjust our model to any
not depend on the parameter T min
e . This means, the condition multiple scattering algorithm. Furthermore, the differences
of long step lengths to implement Bohr’s model is valid for
vanish for integrated or broad beam depth dose curves. They
our purposes. Only the calculation time is affected by T min
e . have only a slight influence on broad beam profile doses in
the Bragg–Peak region.
F. Multiple scattering To determine the radiation length X 0 ( ␳ ) for a material
Protons traversing matter are mainly scattered because of with mass density ␳ in the present voxel we define the ratio:
elastic Coulomb interactions with atomic nuclei. This leads ␳w Xw
to many small-angle deflections. Therefore, the small-angle f X 0共 ␳ 兲 ⫽ 共18兲
␳ X 0共 ␳ 兲
approximation can be applied to the multiple scattering
theory. This leads to a Gaussian distribution22 with a width with the radiation length of water X w ⫽36.0863 cm. The
introduced by Rossi and Greisen:28 crosses in Fig. 3 show this ratio as a function of ␳ for all

冉 冊
materials defined in ICRU Report 46.24 As in EGSnrc,21 the
Es 2
⌬z radiation lengths X w and X 0 ( ␳ ) for elements are calculated
␪ 20 ⫽ . 共17兲
␤p X 0共 ␳ 兲 using Tsai’s formula29 and the radiation lengths for com-
Here, p is the proton momentum and X 0 ( ␳ ) is the radiation pounds are derived from the elemental values. The density
length of the material. E s is a constant parameter, indepen- dependence of f X 0 ( ␳ ) can be represented accurately using
the fit:


dent on proton energy and material composition.
The parameter E s is determined by fitting differential dose 1.19⫹0.44 ln共 ␳ ⫺0.44兲 for ␳ ⭓0.9
distributions as calculated by VMCpro to simulations using
13
GEANT4 in water with nuclear reactions switched off. Using f X 0共 ␳ 兲 ⫽ 1.0446⫺0.2180 ␳ for 0.26⭐ ␳ ⭐0.9, 共19兲
GEANT4 version 5.1 we estimated E s ⫽12.4 MeV. When ver- 0.9857⫹0.0085 ␳ for ␳ ⭐0.26

Medical Physics, Vol. 31, No. 8, August 2004


2267 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2267

FIG. 3. Ratios of the radiation lengths of all materials from ICRU Report 46 FIG. 4. Dose contributions from different reaction channels for 150 MeV
共Ref. 24兲 to the radiation length of water 共crosses兲. The solid line represents protons in water.
the fit of Eq. 共19兲.

if the outliers are neglected as in Eq. 共13兲. The solid line in electromagnetic processes. This can be concluded also from
Fig. 3 shows the fit function. That is, knowledge of the Fig. 4. Here the dose contributions from different reaction
atomic composition in each voxel is unnecessary, mass den- channels for 150 MeV protons in water are shown. This fig-
sity is sufficient for modeling multiple scattering. ure is different to a similar plot presented by Paganetti.31 He
An algorithm for path lengths corrections and lateral de- has classified the dose contributions by the particle type. In
flections is not implemented in VMCpro. The results in Sec. our plot the dose contributions from different reaction chan-
III show that this is not necessary for proton beam therapy nels during nuclear interactions of the primary protons are
dose calculations if step sizes not larger than the voxel reso- presented. For example, the curve labeled with ‘‘neutrons’’ is
lution are used. the contribution of all particles resulting from neutrons pro-
duced in inelastic nuclear reactions of primary protons with
G. Transport of ␦-electrons oxygen nuclei. That is, it is the contribution of particles of
different type created during secondary nuclear interactions
According to Eq. 共4兲 the maximum energy transferable to
of these neutrons. This classification is necessary to estimate
␦-electrons during 250 MeV proton ionization interactions is
the influence of nuclear interaction products of the primary
e ⬇0.6 MeV. This corresponds to a range in water
about T max
protons.
of less than 2.5 mm. This range is smaller for smaller proton
According to ICRU Report 46,24 most of the soft tissue
energies. That is, electron transport can be modeled by the
types consist mainly of the elements hydrogen, carbon, ni-
continuous slowing down approximation 共CSDA兲. For this
trogen, and oxygen. The weight proportions are about 10%
reason, the CSDA range of water calculated using the com-
to 11% hydrogen, 85%–90% carbon plus oxygen and a small
puter tool ESTAR26 is implemented in VMCpro and electron
amount 共up to 5%兲 of nitrogen. According to ICRU Report
stopping power ratio fits comparable to the fits as imple-
63,32 the microscopic nuclear cross sections of protons with
mented in VMC17 are used.
carbon, nitrogen, or oxygen ions, normalized by the atomic
The minimum energy of ␦-electrons is given by the pa-
mass 共to calculate the macroscopic cross sections兲 are in the
rameter T min
e specified by the user. We prefer a value of
same order of magnitude. Therefore, we assume, soft tissue
T min
e ⫽0.1 MeV, corresponding to ranges of less than 0.15
behaves like water in terms of nuclear interactions.
mm in water. This selection provides a reasonable compro-
This is not quite true for various components of the skel-
mise between accuracy and calculation speed. But as shown
eton, because they contain 5%–20% of calcium. The elastic
by Verhaegen and Palmans, the simulation of secondary elec-
and inelastic proton–nucleus cross sections of calcium nor-
trons should not be neglected.30
malized by the atomic mass are about 25% smaller compared
to oxygen. This means, the relative nuclear cross section er-
H. Nuclear interactions
ror can be up to 5% if we also estimate bone tissues by water.
The probability of nuclear reactions compared to ioniza- However, this error should become much smaller if we take
tion interactions is less than 5% for 50 MeV protons with into account all processes, i.e., electromagnetic and nuclear
e ⫽0.1 MeV in water. For larger proton energies this
T min processes. Therefore, the influence of this error to dose dis-
probability becomes smaller, e.g., less than 1% for 200 MeV tributions is assumed to be negligible. Later 共see Sec. III兲 we
protons. Therefore, the influence of nuclear interactions of will show that this assumption is valid for applications in
protons with atomic nuclei can be treated as correction to the proton therapy.

Medical Physics, Vol. 31, No. 8, August 2004


2268 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2268

Because of these assumptions, in VMCpro nuclear inter- function of the incident proton energy T p is tabulated in
actions in human tissue of different type 共including bone兲 are ICRU Report 63. In VMCpro the fit to these data
modeled by nuclear interactions in water. That is, nuclear
cross sections and interaction models are implemented for T O⫽0.65 exp共 ⫺0.0013 T p 兲 ⫺0.71 exp共 ⫺0.0177 T p 兲 ,
共24兲
elastic proton–proton scattering as well as elastic and inelas-
tic proton–oxygen nucleus scattering. is implemented. The real energy of the oxygen nucleus after
elastic collisions T O is sampled from the exponential distri-
bution:

冉 冊
1. Nuclear elastic proton – proton interactions
The macroscopic cross section ⌺ pp (T p ) of proton–proton 1 TO
f 共 T O兲 ⫽ exp ⫺ , with T O⭐T O
max
. 共25兲
collisions in water normalized by the mass density ␳ is cal- TO TO
culated from the microscopic cross section ␴ pp (T p ) accord-
ing to: This distribution is an estimate of the emission spectra tabu-
lated in ICRU Report 63. The range of oxygen ions in water
1 wH is very small, therefore, in VMCpro the energy T O is ab-
⌺ pp 共 T p 兲 ⫽N A ␴ 共 T 兲. 共20兲
␳ A H pp p sorbed locally, contributing in this manner to the dose in the
N A is the Avogadro constant, w H is the weight proportion of present voxel. The direction change of the proton results
hydrogen in water and A H is the atomic weight of hydrogen. from its new energy after the collision.
␴ pp (T p ) can be determined, e.g., by the partial-wave analy-
sis database SAID.33 Instead we use an analytical fit to the 3. Nuclear inelastic proton – oxygen interactions
SAID data between 10 and 300 MeV, i.e., the cross section The macroscopic inelastic cross section ⌺ in(T p ) of pro-
⌺ pp (T p )/ ␳ in cm2 /g is calculated by: tons incident on oxygen nuclei is calculated from the micro-
1 scopic cross section ␴ in(T p ) by:
⌺ 共 T 兲 ⫽0.315 T ⫺1.126⫹3.78 10⫺6 T p , 共21兲
␳ pp p p 1 wO
⌺ in共 T p 兲 ⫽N A ␴ 共 T 兲. 共26兲
with the kinetic proton energy T p in MeV. ␳ A O in p
The angular distribution of protons after the collision can Again ␴ in(T p ) is fitted to ICRU Report 63 data in the energy
also by evaluated using the differential cross sections of range between 7 and 250 MeV leading to:

再 冉 冊
SAID,33 however, in the center-of-mass system this distribu-
tion is almost isotropic. Because of kinematic reasons, this 1 7.85
⌺ in共 T p 兲 ⫽0.001 1.64共 T p ⫺7.9兲 exp ⫺0.064T p ⫹
corresponds to an uniform distribution of the proton energy ␳ Tp


in the laboratory system. Therefore, in VMCpro the kinetic
proton energy after elastic proton–proton interactions is ⫹9.86 . 共27兲
sampled from an uniform distribution. By convention the
proton with smaller energy is called the recoil proton. After Equation 共27兲 becomes negative below 7 MeV. In this case
the collision both protons are tracked like primary protons by the cross section is set to zero.
executing the proton transport algorithm recursively. As a result of inelastic nuclear collisions a variety of dif-
ferent reaction products can arise. Beside secondary protons
2. Nuclear elastic proton – oxygen interactions there are neutrons, deuterons, tritons, alphas and heavier
fragments, photons, electrons, etc. To take all these effects
As in Eq. 共20兲 the total macroscopic elastic cross section into account a complicated algorithm is required like the
⌺ el (T p ) of protons incident on oxygen nuclei is given by: GEANT4 Binary Cascade or pre-compound models.
13,23
How-
1 wO ever, we assume that models of this complexity are not
⌺ el 共 T p 兲 ⫽N A ␴ 共T 兲 共22兲 needed for proton therapy dose calculation.
␳ A O el p
The empirical approach presented here is mainly based on
with w O being the weight proportion of oxygen in water and ICRU Report 6332 and the ENDF database.34 According to
A O being the atomic weight of oxygen. In VMCpro the mi- these tabulations, in inelastic nuclear collisions about 50% of
croscopic cross section ␴ el (T p ) is fitted to data from ICRU the primary kinetic proton energy is released to secondary
Report 6332 in the energy range between 50 and 250 MeV. protons. A significant amount of energy is also released to
This leads to: neutrons, deuterons, alpha particles and heavier nucleus frag-
1 1.88 ments 共see Fig. 4兲. The neutron energy is distributed over
⌺ el 共 T p 兲 ⫽ ⫹4.0 10⫺5 T p ⫺0.014 75. 共23兲 large distances. On the other hand, nuclear fragments have
␳ Tp
short ranges, approximately their energy is absorbed locally.
Again ⌺ el (T p )/ ␳ is in units of cm2 /g and T p is in MeV. Therefore, it is convenient to subdivide the reaction products
max
The maximum energy T O transferable to the recoil into three classes: Protons, short range particles and long
nucleus after the collision is given by Eq. 共4兲 if the electron range particles. Furthermore, binding and minimum energies
mass m e is replaced by the mass of the oxygen nucleus m O . must be taken into account during simulation of inelastic
The average kinetic energy T O of the recoil nucleus as a collisions.

Medical Physics, Vol. 31, No. 8, August 2004


2269 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2269

FIG. 6. Total proton production multiplicity 共upper plot兲 and average sec-
FIG. 5. Secondary proton energy spectrum after inelastic nuclear collisions ondary proton energy 共lower plot兲 as functions of the incident proton energy
of 200 MeV protons incident on Oxygen nuclei. The data represented by the after inelastic nuclear collisions. The data represented by solid lines and
solid line are from ICRU Report 63.32 The dashed line is calculated using circles are from ICRU Report 63.32 The dashed lines and triangles are cal-
the inelastic nuclear model implemented in VMCpro. culated using the inelastic nuclear model implemented in VMCpro.

In the model the energy of secondary particles is sampled The ratio between long range energy and short range en-
from a uniform distribution between the minimum energy ergy is determined by adjusting depth dose curves as calcu-
and the remaining energy of the system consisting of the lated with GEANT4 and VMCpro in water. This leads to 7%
incident proton and the nucleus. At the beginning of this short range energy and 93% long range energy. Short range
iteration loop the remaining energy is given by the energy of energy is deposited locally, long range energy is neglected.
the incident proton minus binding energy. With a probability All secondary protons are handled like primary protons. If
p ⫽0.5 MeV it is absorbed
their energy is smaller than T min
of one half the secondary particle is a proton. Otherwise it is
a short or a long range particle. Then the new remaining locally, otherwise they are tracked by invoking the proton
energy of the system is calculated by subtracting the energy transport procedure recursively. According to ICRU Report
of the first secondary particle and an additional binding en- 63 the angular distribution of secondary protons is forward
ergy. A second secondary particle energy can be sampled if peaked for large energies but approximately isotropic for
there is enough remaining energy, i.e., the iteration loop con- small energies. In VMCpro this is realized by sampling the
tinues. In this manner the loop continues until the whole cosine of the radial scattering angle ␪ s from a uniform dis-
energy of the system is consumed. For 200 MeV protons tribution in the interval:
incident on oxygen nuclei the model leads to a secondary
proton energy spectrum as presented by the dashed line in Ts
2 ⫺1⭐cos ␪ s ⭐1. 共28兲
Fig. 5. The solid line in Fig. 5 shows the corresponding Tp
spectrum from ICRU Report 63. The minimum energy pa- T s is the kinetic energy of the secondary proton, T p is the
rameter of the nuclear model is determined by adjusting the incident proton energy.
low energy peaks of both spectra. This leads to a value of 3
MeV. During the iteration loop the binding energy changes
from 5 MeV to smaller values. As shown in Fig. 5 this leads III. RESULTS AND DISCUSSION
to a maximum secondary proton energy about 5 MeV smaller To verify the results of VMCpro, the general purpose
than the incident energy. Therefore, the overall shape of the Monte Carlo code GEANT413 version 5.2 is used as bench-
spectra is similar. However especially at low energies the mark. GEANT4 is very flexible and allows to choose between
model spectrum is narrower compared to the ICRU data. physical processes and models via the ‘‘physics list.’’ This
Integration of these spectra leads to the total proton produc- provides an easy way to switch on and off nuclear reactions.
tion multiplicities as represented by the upper plot of Fig. 6.
A. Verification of electromagnetic processes
The lower plot shows the average secondary proton energy
as a function of the incident proton energy. Again the solid For verification of electromagnetic processes, elastic and
lines 共with circles兲 are from ICRU Report 63 and the dashed inelastic nuclear reactions are switched off in the following
lines 共with triangles兲 result from the model. In ICRU Report examples. The dose calculations are performed in homoge-
63 the total cross sections are tabulated with an accuracy of neous phantoms consisting of various media. The phantom
5%–10%. The angle-integrated emission spectra are accurate size is 4.1⫻4.1⫻30 cm3 with 1 mm voxel resolution in each
to 20%–30% standard deviation. Taking this into account, direction. Dose distributions of monoenergetic proton pencil
the curves in Fig. 6 are in agreement. beams with different energies are calculated. Pencil beam

Medical Physics, Vol. 31, No. 8, August 2004


2270 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2270

FIG. 7. Integral depth dose distributions in adult soft tissue ( ␳ FIG. 8. Central axis depth dose distributions in adult soft tissue
⫽1.03 g/cm3 ) and adult skeleton ( ␳ ⫽1.46 g/cm3 ) for monoenergetic ( ␳ ⫽1.03 g/cm3 ) and adult skeleton ( ␳ ⫽1.46 g/cm3 ) for monoenergetic
proton pencil beams of 100 MeV and 200 MeV, calculated with GEANT4 and proton pencil beams of 100 MeV and 200 MeV, calculated with GEANT4 and
VMCpro. Nuclear reactions are switched off in both MC codes. VMCpro. Nuclear reactions are switched off in both MC codes.

dose distributions are more sensitive compared to broad


results of GEANT4 simulations without and with nuclear re-
beams with respect to the model assumptions and the behav-
actions as well as FLUKA with nuclear reactions. It shows that
ior of the MC code. Therefore, broad beam results are not
these processes cannot be neglected for dose calculation.
presented here. But it is possible to integrate pencil beam
This result is consistent with experiences of other research
dose distributions perpendicular to the beam axis. The result-
groups.31,35 The nuclear reactions seem to be more important
ing dose distributions are equivalent to broad beam depth
for high proton energies 共200 MeV兲 compared to lower en-
dose curves.
ergies 共100 MeV兲. This is mainly an accumulative effect,
Integrated depth dose curves in adult soft tissue and adult
because for larger proton energies and for larger penetration
skeleton as defined in ICRU Report 4624 for two typical en-
depths there is an increased chance of nuclear interactions.
ergies in proton beam therapy 共100 and 200 MeV兲 are shown
Thus, the height of the Bragg–Peak becomes smaller.
in Fig. 7. Slight differences between both MC codes can be
Figure 10 shows the same comparison between GEANT4
detected only near the Bragg–Peak. The maximum devia-
and VMCpro as Fig. 7, however, this time all nuclear inter-
tions are smaller than 1% or 0.5 mm. The differences are
actions are switched on in both MC codes. The maximum
larger 共especially the proton range兲 if we use standard
GEANT4 proton stopping powers. To have identical conditions
we employed the low energy extension with ICRU – R49p
option in GEANT4.
Differential depth dose curves through the central voxels
for the same beam and phantom configurations are shown in
Fig. 8. These curves are sensitive especially with respect to
the implementation of multiple scattering. It shows that the
proton transport algorithm with small-angle multiple scatter-
ing approximation, without path length correction and with-
out lateral deflections works well for this type of applica-
tions.
Figures 7 and 8 demonstrate also the accuracy of the den-
sity scaling functions Eq. 共13兲 and Eq. 共19兲. Only the mass
density is used as input for VMCpro. GEANT4, on the other
hand, takes into account the real atomic composition. Espe-
cially the dose calculations in a phantom consisting only of
bone tissue are extremely unrealistic. Nevertheless, the
agreement is quite good.

B. Verification of nuclear processes FIG. 9. Integral depth dose distributions in water for monoenergetic proton
pencil beams of 100 MeV and 200 MeV, calculated using GEANT4 with
To motivate the implementation of elastic and inelastic nuclear reactions switched off and on as well as FLUKA with nuclear reac-
nuclear collisions into VMCpro Fig. 9 is shown. It represents tions.

Medical Physics, Vol. 31, No. 8, August 2004


2271 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2271

FIG. 10. Same as Fig. 7, but this time nuclear reactions are switched on in FIG. 12. Central axis depth dose distributions 共upper plot兲 in the phantom
both MC codes. consisting of water, bone and lung tissue 共see Fig. 11兲 for monoenergetic
proton pencil beams of 150 MeV, calculated with GEANT4 and VMCpro. The
lower plot shows the difference of both curves in % of the dose maximum.

deviations between the two algorithms are still smaller than


1% or 0.5 mm. Especially the agreement in skeleton demon-
strates that precise knowledge of the atomic composition is
ment of both curves is acceptable, although there is a slight
not necessary to correctly model nuclear processes for treat-
overestimation 共see lower plot兲 of the dose calculated with
ment planning purposes. The approximation of proton–
VMCpro between 5 and 15 cm depth. This overestimation
nucleus collisions by nuclear proton–hydrogen and proton–
can also be observed in the profile plots of Fig. 13, especially
oxygen collisions is sufficient.
in 13 cm depth. The reason is, VMCpro underestimates the
multiple scattering angle slightly in lung. Only the mass den-
C. Verification using an inhomogeneous phantom sity of lung (0.26 g/cm3 ) is used as input for VMCpro. On
For verification in inhomogeneous geometries we con- the other hand, there is still a problem with multiple scatter-
struct a water phantom of size 4⫻4⫻30 cm3 , again with 1 ing in GEANT4, because different program versions produce
mm voxel resolution 共see Fig. 11兲. The phantom contains different results 共see Fig. 2兲. These differences vanish for
two rectangular inhomogeneities of size 1⫻2⫻5 cm3 . They broader, clinically more realistic, proton beams or laterally
consist of bone and lung tissue and are placed starting at 5 integrated dose distributions 共see Fig. 14兲.
cm depth. The inhomogeneities are located adjacent to each
other with the interface between them being on the pencil
beam axis, i.e., the whole inhomogeneity is of size 2⫻2
⫻5 cm3 .
The central axis depth dose distribution of a 150 MeV
pencil beam is shown in Fig. 12. Because in this case there is
no central voxel with respect to x and y as in the homoge-
neous cases, the presented curve is averaged over the four
central voxels. This leads to smaller doses 共by a factor of
about four兲, because the pencil beam spreads energy equally
to four voxels instead of one. The positions of both Bragg–
Peaks, resulting from the inhomogeneities, are well repro-
duced by VMCpro compared to GEANT4. The overall agree-

FIG. 11. Phantom consisting of water, bone and lung tissue for the bench-
mark calculations. The proton pencil beam of 150 MeV hits the inhomoge- FIG. 13. Profiles in depths of 13 cm and 18 cm for the example of Figs. 11
neity at the interface between bone and lung. and 12.

Medical Physics, Vol. 31, No. 8, August 2004


2272 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2272

is a 100 MeV 5⫻5 cm2 proton beam incident on a human


head. Besides the fact that this is not a clinical case,
the influence of air cavities and bone structures on the dose
distribution can be clearly observed. The voxel resolution
is 2⫻2⫻2 mm3 and 500 000 histories are simulated. The
calculation time is 106 seconds using an AMD Athlon
PC 2400⫹. The statistical variance is about 1.5% of the
maximum dose.

E. Calculation time comparison


The inhomogeneous phantom example of Sec. III C is
also used to compare the calculation times of the different
MC codes. To compute the dose distributions one half mil-
lion histories are simulated using an AMD Athlon PC 2400
⫹. For dose calculation, GEANT4 is used with low energy
extension. This was necessary to have the proton stopping
powers from ICRU Report 49. However, this slows down the
FIG. 14. Integral depth dose distributions for the example of Figs. 11 and 12. computation time by about 30%. GEANT4 共with low energy
extension兲, GEANT4 共without low energy extension兲, FLUKA
and VMCpro—all with identical electron cut-off energies—
D. Dose calculation based on CT images require 99 min, 65 min, 37 min, and 168 s CPU time, respec-
The conversion functions Eqs. 共13兲 and 共19兲 as well as tively. That is, VMCpro is about 35 times faster than GEANT4
the implementation of nuclear interactions in VMCpro al- with low energy extension, about 23 times faster than
lows dose calculations based on mass density distributions. GEANT4 without low energy extension and 13 times faster
That is, knowledge of the atomic composition in each than FLUKA. The CPU time reduction is larger for dose cal-
voxel is not required using this method. Mass density is a culations in patients, because VMCpro calculation times are
material parameter. It is not dependent on CT scanner ener- similar in homogeneous phantoms and heterogeneous geom-
gies like Hounsfield numbers. However, mass density can be etries. This is not the case for GEANT4 and FLUKA. The re-
determined if the CT is calibrated appropriately. Figure 15 quirement of a large material data base for the cross sections
shows an example of dose calculation using VMCpro. It makes simulations in patient geometries slower.

IV. CONCLUSIONS
In the present paper a Monte Carlo algorithm for treat-
ment planning in proton beam therapy of cancer is intro-
duced. The algorithm, called VMCpro, is faster than the gen-
eral purpose MC codes FLUKA and GEANT4 by at least one
order of magnitude for simulations in a water phantom with
inhomogeneities. For dose calculations in patients the speed
improvement is larger. The results produced with VMCpro
are in agreement with the GEANT4 results. For integrated or
broad beam depth dose curves the maximum deviations are
not larger than 1% or 0.5 mm in regions with large dose
gradients. For differential pencil beam dose distributions the
deviation can be larger, because these representations are
generally more sensitive compared to broad beam dose dis-
tributions. However, normalized to the dose maximum, the
agreement is still within 2%.
VMCpro shows larger deviations compared to FLUKA.
There are also deviations in the results of different versions
of GEANT4 or between GEANT4 and FLUKA. This is mainly
due to various multiple scattering distributions in these mod-
els. Therefore, further research will be necessary to deter-
mine, which of the distributions is correct. Fortunately,
VMCpro can be adjusted to any model. The multiple scatter-
FIG. 15. Dose distribution of a 5⫻5cm2 proton beam of 100 MeV in a
ing distribution of VMCpro is fitted to GEANT4 version 5.2.
patient’s head. The spatial resolution of the density and dose grids is 2⫻2 Thus the deviations between VMCpro and GEANT4 version
⫻2 mm3 . 5.2 are much smaller than the deviations between the differ-

Medical Physics, Vol. 31, No. 8, August 2004


2273 M. Fippel and M. Soukup: A Monte Carlo dose calculation algorithm 2273

15
ent general purpose Monte Carlo codes. A second reason for A. Fasso, A. Ferrari, J. Ranft, and P. R. Sala, ‘‘FLUKA: Status and Pro-
spective for Hadronic Applications,’’ in Advanced Monte Carlo for Ra-
disagreement between VMCpro and FLUKA are different
diation Physics, Particle Transport Simulation and Applications, Pro-
stopping power implementations. VMCpro is based on ICRU ceedings of the Monte Carlo 2000 Conference, Lisbon, October 23–26,
Report 49,25 therefore, especially in bone tissue the proton 2000, edited by A. Kling, F. Barao, M. Nakagawa, L. Tavora, and P. Vaz
ranges can be different by about 1 mm. In GEANT4 we can 共Springer-Verlag, Berlin, 2000兲, pp. 955–960.
16
switch on the ICRU – R49p option to be conform to ICRU Re- J. F. Briesmeister, MCNP—a General Monte Carlo N-Particle Transport
Code, Los Alamos National Laboratory 共1997兲, Report No. LA-12625-M.
port 49. 17
I. Kawrakow, M. Fippel, and K. Friedrich, ‘‘3D electron dose calculation
using a Voxel based Monte Carlo algorithm 共VMC兲,’’ Med. Phys. 23,
ACKNOWLEDGMENTS 445– 457 共1996兲.
18
M. Fippel, ‘‘Fast Monte Carlo dose calculation for photon beams based
Financial support for the work presented here was pro-
on the VMC electron algorithm,’’ Med. Phys. 26, 1466 –1475 共1999兲.
vided by the Deutsche Forschungsgemeinschaft. One of us 19
I. Kawrakow and M. Fippel, ‘‘Investigation of variance reduction tech-
共M. Soukup兲 is supported by a Marie Curie Fellowship of the niques for Monte Carlo photon dose calculation using XVMC,’’ Phys.
European Commission. We would like to thank Johannes- Med. Biol. 45, 2163–2183 共2000兲.
20
Peter Wellisch from CERN for his suggestions regarding the M. J. Berger, Monte Carlo Calculation of the penetration and diffusion of
fast charged particles, Methods in Computational Physics, edited by B.
nuclear interaction models in GEANT4. Alder, S. Fernbach, and M. Rotenberg 共Academic, New York, 1963兲, Vol.
I, pp. 135–215.
a兲
Electronic mail: msfippel@med.uni-tuebingen.de 21
I. Kawrakow, ‘‘Accurate condensed history Monte Carlo simulation of
1
P. Petti, ‘‘Differential-pencil-beam dose calculations for charged par- electron transport, I. EGSnrc, the new EGS4 version,’’ Med. Phys. 27,
ticles,’’ Med. Phys. 19, 137–149 共1992兲. 485– 498 共2000兲.
2
M. Lee, A. E. Nahum, and S. Webb, ‘‘An empirical method to build up a 22
K. Hagiwara et al., ‘‘Review of Particle Physics,’’ Phys. Rev. D 66,
model of proton dose distribution for a radiotherapy treatment-planning 010001 共2002兲, available on Particle Data Group WWW pages 共http://
package,’’ Phys. Med. Biol. 38, 989–998 共1993兲.
3 pdg.lbl.gov兲.
K. R. Russell, E. Grusell, and A. Montelius, ‘‘Dose calculations in proton 23
GEANT4 collaboration, Physics Reference Manual 共2003兲, available at
beams: Range straggling corrections and energy scaling,’’ Phys. Med.
http://wwwasd.web.cern.ch/wwwasd/geant4/geant4.html.
Biol. 40, 1031–1043 共1995兲. 24
4 ‘‘Photon, Electron, Proton and Neutron Interaction Data for Body Tis-
L. Hong, M. Goitein, M. Bucciolini, R. Comiskey, B. Gottschalk, S.
Rosenthal, C. Serago, and M. Urie, ‘‘A pencil beam algorithm for proton sues,’’ ICRU-Report 46, International Commission on Radiation Units
dose calculations,’’ Phys. Med. Biol. 41, 1305–1330 共1996兲. and Measurements 共1992兲.
25
5
A. K. Carlsson, P. Andreo, and A. Brahme, ‘‘Monte Carlo and analytical ‘‘Stopping Powers and Ranges for Protons and Alpha Particles,’’ ICRU-
calculation of proton pencil beams for computerized treatment plan opti- Report 49, International Commission on Radiation Units and Measure-
mization,’’ Phys. Med. Biol. 42, 1033–1053 共1997兲. ments 共1993兲.
26
6
J. O. Deasy, ‘‘A proton dose calculation algorithm for conformal therapy M. J. Berger, ‘‘ESTAR, PSTAR, and ASTAR: Computer Programs for Calcu-
simulations based on Moliere’s theory of lateral deflections,’’ Med. Phys. lating Stopping-Power and Range Tables for Electrons, Protons, and He-
25, 476 – 483 共1998兲. lium Ions,’’ Technical Report NBSIR 4999 National Institute of Standards
7
G. A. Sandison and A. V. Chvetsov, ‘‘Proton loss model for therapeutic and Technology, Gaithersburg, MD 共1993兲.
beam dose calculations,’’ Med. Phys. 27, 2133–2145 共2000兲. 27
N. Bohr, ‘‘The penetration of atomic particles through matter,’’ K. Dan.
8
H. Szymanowski, A. Mazal, C. Nauraye, S. Biensan, R. Ferrand, M. C. Vidensk. Selsk. Mat. Fys. Medd. 18共8兲, 1–144 共1948兲.
Murillo, S. Caneva, G. Gaboriaud, and J. C. Rosenwald, ‘‘Experimental 28
B. Rossi and K. Greisen, ‘‘Cosmic-ray theory,’’ Rev. Mod. Phys. 13,
determination and verification of the parameters used in a proton pencil 240–309 共1941兲.
beam algorithm,’’ Med. Phys. 28, 975–987 共2001兲. 29
Y. S. Tsai, ‘‘Pair production and bremsstrahlung of charged leptons,’’
9
A. Lomax, ‘‘Intensity modulation methods for proton radiotherapy,’’ Rev. Mod. Phys. 46, 815– 851 共1974兲.
Phys. Med. Biol. 44, 185–205 共1999兲. 30
F. Verhaegen and H. Palmans, ‘‘A systematic Monte Carlo study of sec-
10
B. Schaffner, E. Pedroni, and A. Lomax, ‘‘Dose calculation models for ondary electron fluence perturbation in clinical proton beams 共70–250
proton treatment planning using a dynamic beam delivery system: an MeV兲 for cylindrical and spherical ion chambers,’’ Med. Phys. 28, 2088 –
attempt to include density heterogeneity effects in the analytical dose 2095 共2001兲.
calculation,’’ Phys. Med. Biol. 44, 27– 41 共1999兲. 31
H. Paganetti, ‘‘Nuclear interactions in proton therapy: Dose and relative
11
H. Szymanowski and U. Oelfke, ‘‘Two-dimensional pencil beam scaling:
biological effect distributions originating from primary and secondary
an improved proton dose algorithm for heterogeneous media,’’ Phys.
particles,’’ Phys. Med. Biol. 47, 747–764 共2002兲.
Med. Biol. 47, 3313–3330 共2002兲. 32
12 ‘‘Nuclear Data for Neutron and Proton Radiotherapy and for Radiation
R. Kohno, Y. Takada, T. Sakae, T. Terunuma, K. Matsumoto, A. Nohtomi,
Protection,’’ ICRU-Report 63, International Commission on Radiation
and H. Matsuda, ‘‘Experimental evaluation of validity of simplified
Monte Carlo method in proton dose calculations,’’ Phys. Med. Biol. 48, Units and Measurements 共2000兲.
33
1277–1288 共2003兲. R. A. Arndt, I. I. Strakovsky, and R. L. Workman, ‘‘Nucleon–nucleon
13
S. Agostinelli et al., ‘‘GEANT4—a simulation toolkit,’’ Nucl. Instrum. elastic scattering to 3 GeV,’’ Phys. Rev. C 62, 034005 共2000兲.
34
Methods Phys. Res. A 506, 250–303 共2003兲. V. McLane, ‘‘ENDF-102 Data Formats and Procedures for the Evaluated
14
A. Fasso, A. Ferrari, and P. R. Sala, ‘‘Electron–photon transport in Nuclear Data File ENDF-6,’’ Technical Report BNL-NCS-44945-01/04-
FLUKA: status,’’ in Advanced Monte Carlo for Radiation Physics, Par- Rev. Brookhaven National Laboratory, National Nuclear Data Center,
ticle Transport Simulation and Applications, Proceedings of the Monte Upton, NY, USA 共2001兲.
35
Carlo 2000 Conference, Lisbon, October 23–26, 2000, edited by A. J. Medin and P. Andreo, ‘‘Monte Carlo calculated stopping-power ratios,
Kling, F. Barao, M. Nakagawa, L. Tavora, and P. Vaz 共Springer-Verlag, water–air, for clinical proton dosimetry 共50–250 MeV兲,’’ Phys. Med.
Berlin, 2000兲, pp. 159–164. Biol. 42, 89–105 共1997兲.

Medical Physics, Vol. 31, No. 8, August 2004

You might also like