NIH Public Access

Author Manuscript
Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
Published in final edited form as:
NIH-PA Author Manuscript

Phys Med Biol. 2012 June 7; 57(11): R99–R117. doi:10.1088/0031-9155/57/11/R99.

Range uncertainties in proton therapy and the role of Monte

Carlo simulations
Harald Paganetti
Department of Radiation Oncology, Massachusetts General Hospital and Harvard Medical
School, Boston, MA 02114

Abstract
The main advantages of proton therapy are the reduced total energy deposited in the patient as
compared to photon techniques and the finite range of the proton beam. The latter adds an
additional degree of freedom to treatment planning. The range in tissue is associated with
considerable uncertainties caused by imaging, patient setup, beam delivery and dose calculation.
Reducing the uncertainties would allow a reduction of the treatment volume and thus allow a
NIH-PA Author Manuscript

better utilization of the advantages of protons. This article summarizes the role of Monte Carlo
simulations when aiming at a reduction of range uncertainties in proton therapy. Differences in
dose calculation when comparing Monte Carlo with analytical algorithms are analyzed as well as
range uncertainties due to material constants and CT conversion. Range uncertainties due to
biological effects and the role of Monte Carlo for in vivo range verification are discussed.
Furthermore, the current range uncertainty recipes used at several proton therapy facilities are
revisited. We conclude that a significant impact of Monte Carlo dose calculation can be expected
in complex geometries where local range uncertainties due to multiple Coulomb scattering will
reduce the accuracy of analytical algorithms. In these cases Monte Carlo techniques might reduce
the range uncertainty by several mm.

1. Introduction
The number of patients being treated with proton therapy is steadily increasing in radiation
oncology. The main reason for the increased interest in proton therapy is the physical
characteristics of the depth dose curve with a dose peak (Bragg peak) at a well-defined depth
in tissue, which was first recognized as potentially useful for cancer therapy by Wilson
NIH-PA Author Manuscript

(Wilson, 1946). The total energy deposited in a patient (often termed ‘integral dose’) for a
given target dose is always lower when using proton as compared to photon treatment
techniques. This is in part due to the lack of exit dose. Furthermore, the Bragg peak allows,
in theory, to point a beam directly towards a critical structure, increasing the flexibility of
proton therapy treatment planning compared to conventional photon therapy. A short
overview of proton therapy delivery techniques is given elsewhere (Paganetti and Kooy,
2010).

In order to fully utilize the potential advantage when using protons, the range of proton
beams in patients needs to be predicted as accurate as possible in the treatment planning and
delivery process. An improper quantification of safety margins can have more sever
consequences in proton therapy than in photon therapy. Margins underestimating
uncertainties in photon therapy might cause under-dosage of the tumor. In proton therapy,
such an underestimation may cause part of the tumor not receiving any dose due to a
potential shift of the sharp distal dose fall-off.
 Paganetti Page 2

Uncertainties in the exact position of the distal dose gradient arise from a) organ motion, b)
setup and anatomical variations, c) dose calculation approximations, and d) biological
considerations. At the Massachusetts General Hospital (MGH), treatment planning assumes
NIH-PA Author Manuscript

an uncertainty in the proton beam range of 3.5% of the range plus an additional 1mm. Other
centers follow similar margin recipes. For example, the MD Anderson Proton Therapy
Center in Houston, the Loma Linda University Medical Center, and the Roberts Proton
Therapy Center at the University of Pennsylvania all apply 3.5% + 3mm while the
University of Florida Proton Therapy Institute uses 2.5% + 1.5mm1. The 3.5% + 1mm range
margin at MGH, for instance, results in a deliberate overshoot of 8mm for a 20 cm range
field in soft tissue, which is quite substantial. Note that these margins are often not fully
generic and that adjustment may be made for certain sites based on the location of critical
structures.

Monte Carlo techniques are being used in proton therapy for years. However, the improved
accuracy of Monte Carlo as compared to analytical dose calculation techniques had no
influence on the margin recipes so far. Comparisons of the two dose calculation techniques
reveal in general differences that are clinically insignificant because conservative treatment
margins minimize the impact of dose calculation (e.g. range) uncertainties. Progress in
radiation therapy and decrease in normal tissue complication probability can be achieved in
part through the reduction of treatment volumes. Only if we understand range uncertainties
will we feel comfortable to reduce the range margin. As part of an international workshop
NIH-PA Author Manuscript

on “Recent Advances in Monte Carlo Techniques for Radiation Therapy” we tried to answer
the question whether Monte Carlo dose calculation might be able to contribute to this quest.
This article summarizes the presentation given at this meeting. It does focus on static cases
without organ motion.

2. Range definition
Due to range straggling, not all protons of the same energy have the same range. Thus, the
range needs to be defined for a beam of protons resulting in a broadened Bragg peak or a
spread-out Bragg peak (SOBP). Ideally the range would be defined at the position where the
dose has decreased to 80% of the maximum dose, i.e. in the distal dose falloff. The reason
for this choice is the fact that for a monoenergetic proton beam, the 80% fall-off position
coincides with the mean projected range of a proton, i.e. the range at which 50% of the
protons have stopped. Furthermore, the 80% fall-off position is thus independent of the
beam’s energy spread. Nevertheless, in most proton therapy facilities, the prescribed range is
defined at the 90% fall-off position in water or the 90% isodose line in a patient because of
historic reasons. Figure 1 shows the energy-range relationship for monoenergetic proton
beams.
NIH-PA Author Manuscript

3. Studies on range uncertainties in proton therapy

3.1 Range uncertainties due to multiple Coulomb scattering
While Monte Carlo based treatment planning systems are slowly entering the market, most
current commercial planning systems are based on analytical pencil-beam dose calculation
algorithms. Differences between Monte Carlo based dose calculation and analytical methods
in proton therapy have been demonstrated (Paganetti et al., 2008; Petti, 1996; Schaffner et
al., 1999; Soukup and Alber, 2007; Soukup et al., 2005; Szymanowski and Oelfke, 2002;
Tourovsky et al., 2005; Titt et al., 2008a). In addition, experimental validations have been

1Personal communications by Stephen Avery (Roberts Proton Therapy Center at the University of Pennsylvania), Zuofeng Li
(University of Florida Proton Therapy Institute), Radhe Mohan (MD Anderson Proton Therapy Center, Houston), and Baldev Patyal
(Loma Linda University Medical Center).

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 3

done in simple geometries. If we consider Monte Carlo to be the gold standard (although
there are physics uncertainties in the Monte Carlo as well; see below) we can assess the
range uncertainty due to dose calculation in complex geometries by comparing the range
NIH-PA Author Manuscript

predicted by the treatment planning system with the range predicted by a Monte Carlo
simulation.

Analytical dose calculation algorithms typically project the range based on the water
equivalent depth in the patient neglecting the position of inhomogeneities relative to the
Bragg peak depth (Petti, 1992, 1996; Urie et al., 1984). Furthermore, such algorithms are
less sensitive to complex geometries and density variations, e.g. at bone-soft tissue
interfaces. Consequently, analytical algorithms are not able to correctly predict the effect of
range degradation caused by multiple Coulomb scattering (Sawakuchi et al., 2008; Urie et
al., 1986; Goitein, 1977; Goitein and Sisterson, 1978). Figure 2 illustrates the effect in a
regular geometrical grid with alternating high and low density material based on a Monte
Carlo study (Sawakuchi et al., 2008). The authors analyzed range degradation effects as a
function of proton energy and geometric complexity. It was found that the difference in the
80% to 20% falloff position is in the order of ~2mm for a 220 MeV beam. At the 90%
falloff the figure shows a difference in range of about 1.5mm, i.e. −0.7%.

While in such a regular structure range degradation causes mainly a simple widening of the
distal fall-off, the effects in a complex patient geometry are not as easy to predict. For an
NIH-PA Author Manuscript

SOBP penetrating a human skull, it was reported that dose degradation in the distal falloff
can reduce the range at the 80% dose level by up to 8mm, depending on the density
variation (Urie et al., 1986). This worst case did affect only a small region of the distal dose
distribution. Another example is shown in figure 3. Here, multiple Coulomb scattering at a
bone-soft tissue interface tangential to the beam, in the penumbra, causes a scatter dis-
equilibrium. The result is a range variation affecting a small part of the distal fall-off only.
This effect can be understood considering the simple geometry in figure 4. Range
perturbations due to multiple Coulomb scattering in complex structures cannot be modeled
using solely the stopping power of materials involved and their water equivalent path length,
which would simply cause a shift in the distal falloff.

While for large fields these might be local effects, for very small fields in the head and neck
region these effects can potentially affect the majority of the treatment volume as seen in
figure 5 (Bednarz et al., 2010). The figure illustrates a dose non-uniformity in the target
predicted by Monte Carlo but not by the treatment planning system. The effect of range
degradation as illustrated by the step in the distal part of the depth dose distribution is quite
substantial on the 95% dose level but almost negligible at the 90% level. The example is for
a small field with complete lack of lateral equilibrium and thus not a typical scenario. For
NIH-PA Author Manuscript

these, the range effects seen in figures 3 and 5 can be estimated to be on the order of ±2.5%.
Note that the dose discrepancy near the entrance region is due to scattering from the aperture
located close to the patient, which is not considered by the analytical algorithm. The impact
of edge-scattered protons has been studied previously (Titt et al., 2008b; van Luijk et al.,
2001). Not modeled accurately in analytical dose calculations will affect mainly the
proximal part of the Bragg curve, except for very shallow fields.

3.2 Monte Carlo uncertainties due to physics

In the previous section we have shown that analytical dose calculations do have limitations
when it comes to predicting the correct proton beam range. However, there are also
uncertainties when using Monte Carlo methods. Three codes used in proton therapy (Geant4
(Agostinelli et al., 2003), FLUKA (Ferrari et al., 2005), and MCNPX (Pelowitz, 2005)) were
compared with each other and with measurements by Kimstrand et al (Kimstrand et al.,
2008). The physics settings within the codes were varied when simulating small angle

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 4

scattering at aperture edges where small uncertainties in scattering power would be

detectable. A significant impact of user-defined parameters was found. One has to keep in
mind that most Monte Carlo codes used in proton therapy were originally developed for
NIH-PA Author Manuscript

high-energy physics applications and span a wide range of particles and energy domains.
There can be uncertainties due to the implementation of physics interactions via theoretical
models or interpolation of experimental data. The accuracy of physics settings might depend
on the energy region of interest. Thus, the default physics models might not be tailored for
proton therapy simulations and may require adjustment (Herault et al., 2005; Stankovskiy et
al., 2009; Pia et al.,2010; Zacharatou Jarlskog and Paganetti, 2008). It is the responsibility of
Monte Carlo users to benchmark their code and understand the impact of incorrect physics
settings. In particular, uncertainties in nuclear interaction probabilities can be substantial
because most experiments were done decades ago on thin targets resulting in data for one
specific energy only. Thus, most cross sections are known over the entire energy region of
interest. Fortunately, nuclear cross section uncertainties are largely inconsequential for the
predicated range from dose calculations because of the low proton energies (below the
nuclear interaction threshold) at the distal falloff of a Bragg curve.

A validation of Monte Carlo dose calculation methods is typically done using homogeneous
or heterogeneous geometries consisting of various materials (Paganetti et al., 2004;
Paganetti et al.,2008; Titt et al., 2008a). Beam profiles in water downstream of
inhomogenous targets and compensators have been used to analyse the multiple scattering
NIH-PA Author Manuscript

theory can vary from Moliére’s theory (Andreo et al., 1993; Gottschalk et al.,1993). For
instance, the code Geant4 uses a condensed history algorithm for multiple scattering that
utilizes functions to calculate the angular and spatial distributions of the scattered particle
implementations (Urban, 2002).

While discrepancies between different multiple Coulomb scattering implementations might

be negligible, the uncertainties in the mean excitation energy are presumably significant.
The two main parameters in the Bethe-Bloch equation that determine the range are the
density and the mean excitation energy. In its report from 1993, the ICRU (ICRU, 1993)
recommends a mean excitation energy, or I-value, of 75±3 eV for water. Others have
reported 80±2 eV (Bichsel and Hiraoka, 1992). The stopping power tables by Janni (Janni,
1982) are based on a value of 81.8 eV while a more recent experimental study resulted in a
value of 78.4±1.0 eV (Kumazaki et al., 2007). Andreo (Andreo, 2009) has demonstrated that
a variation of I-values between 67.2 eV and 80 eV can have a substantial impact on the
proton beam range in water (see figure 6). It was concluded that the range uncertainty due to
uncertainties in I-values is on the order of ±1.5–2.0%. However, the lower boundary can
most certainly be excluded based on recent measurements. The difference in I-value
between 75 and 80 eV results in a 0.8–1.2% difference in the predicted stopping power in
NIH-PA Author Manuscript

the therapeutic energy range.

Even if the Monte Carlo predicted range is normalized to the expected range based on
measurements in water, there are still uncertainties in tissues based on their material
compositions. Mean excitation energies for various elements are tabulated by the ICRU
(ICRU, 1993), which also includes averaged values for tissues (ICRU, 1989, 1992). Mean
excitation energies are typically adjusted to agree with measurement or interpolated where
such data are not available. Theoretical models have uncertainties, for example due to
binding effects (Ahlen, 1980). For tissues, the uncertainty in I-value is potentially on the
order of 10–15% (Andreo, 2009). This might result in a range uncertainty in tissue of
~1.5%.

For complex passive scattered delivery systems, Monte Carlo dose calculation is often based
on a complete simulation of the treatment head geometry. A SOBP is created by varying the

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 5

beam energy by using a rotating absorber consisting of several steps of water-equivalent

thicknesses. The beam current is typically modulated as a function of the wheel rotational
angle. A double-scattering system creates a broad beam. Patient specific apertures and
NIH-PA Author Manuscript

compensators ensure lateral and distal target conformation. Figure 7 shows a simulated
proton therapy treatment head. There are several uncertainties in addition to I-values in
treatment head modeling that can cause Bragg curves to differ slightly from measured ones.
The vendor’s blueprints may deviate slightly from the built devices. Furthermore,
uncertainties are introduced because material compositions might not be known with
sufficient accuracy. For example, Carbon used in scattering devices can be purchases in
various different densities.

A comprehensive study on the uncertainties in treatment head modeling was done by

Bednarz et al. (Bednarz et al., 2011). Specifically the sensitivity of SOBP dose distributions
on material densities, mean excitation energies, initial proton beam energy spread and spot
size was investigated. Densities were varied by up to ±5% and I-values were varied by up to
±20%. For a treatment head simulation based on the MGH gantry (Paganetti et al., 2004)
there was a 4% range uncertainty due to uncertainties in densities of lead and lexan.
Furthermore, the mean excitation energies caused an uncertainty of up to 2%. Note that the
parameter variations may be too generous depending on the quality of the vendor
information.
NIH-PA Author Manuscript

Some of the uncertainties in treatment head modeling cannot be overcome because exact
geometrical or material information from the vendor might not be available. On the other
hand adjusting parameters to match a specific range and modulation width can have a
negative impact on other range/modulation width combinations. An alternative method to
fine-tune a treatment head model is to optimize the time-dependent beam current modulation
functions to improve the agreement between calculated and measured SOBP dose
distribution to within clinical tolerance levels (Bednarz et al., 2011).

3.3 Monte Carlo uncertainties due to CT conversion

In proton therapy planning systems, Hounsfield unit (HU) conversion schemes are
implemented to translate HU into relative stopping power. Conversion schemes are typically
based on CT scans of tissue phantom materials with known density and elemental
composition. The relationship between stochastic errors, i.e. noise, in CT images and the
predicted proton beam range has been studied by Chvetsov and Paige (Chvetsov and Paige,
2010). They reported a standard deviation of the range between ~0.3% and ~0.7% assuming
a CT grid of 3mm and a 2.5% noise level. An error below 1% on the stopping power values
was reported based on dual energy CT (Yang et al., 2010).
NIH-PA Author Manuscript

Even small discrepancies in the calculated local energy deposition can result in significant
changes in range because they may accumulate over the entire beam path (Jiang and
Paganetti, 2004; Matsufuji et al., 1998; Mustafa and Jackson, 1983). A range uncertainty
margin of 0.82% was suggested by Matsufuji et al. (Matsufuji et al., 1998). Density
variations alone might cause an uncertainty below 1% because of the small spread of
densities for tissues corresponding to the same HU. The range uncertainty of the conversion
method based on a stoichiometric conversion using animal tissues has been reported to be up
to 1.8% for bone and 1.1% for soft tissues translating into a range uncertainty of 1–3mm
(Schaffner and Pedroni, 1998). These numbers do include the uncertainty in the I-value.
Thus, the pure uncertainty due to stopping power conversion is presumably ~0.5%.

For Monte Carlo simulations, not relative stopping power but density and material
composition are assigned. Various approaches of material/density conversions have been
published (du Plessis et al., 1998; Schneider et al., 1996; Schneider et al., 2000;

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 6

Vanderstraeten et al., 2006). The stopping power is only slightly affected by uncertainties in
the carbon, oxygen and nitrogen weight fractions because they have very similar atomic
number to atomic mass ratio. It is more sensitive to hydrogen weight fraction. On the other
NIH-PA Author Manuscript

hand, the uncertainty of hydrogen weight fraction obtained from a stoichiometric calibration
is negligible for most tissues (Schneider et al., 2000). Consequently, CT conversion
uncertainties using Monte Carlo should be smaller than the ones for analytical dose
calculation. It has been shown that if the HU conversion is based on a proper stoichiometric
calibration, the uncertainties on the proton beam range for head and neck patients is likely
below 0.5mm (~0.2%) (Espana and Paganetti, 2010).

While each HU is assigned a unique density, the number of material composition

assignments is typically much smaller. Fewer tissue (i.e. material compositions) do increase
the efficiency of Monte Carlo simulations as physics settings are typically generated for
each tissue separately on the fly. It was shown that the number of different tissues defined in
a Monte Carlo simulation should be larger than ~10 to avoid an impact on the proton beam
range (Espana and Paganetti, 2010).

Note that when using a conversion scheme for clinical dose calculation it has to be
normalized to the departmental CT scanner. This can be done, for example, by simulating a
HU versus relative stopping power conversion curve in the Monte Carlo setting and then
normalize this to the clinical conversion curve by slightly adjusting the HU versus density
NIH-PA Author Manuscript

relationship in the Monte Carlo (Jiang and Paganetti, 2004).

Potential CT artifacts can also affect the accuracy of proton range prediction based on CT
images. These are particularly severe in the presence of metallic implants (Jaekel and Reiss,
2007; Newhauser et al., 2008). We have not included this specific scenario in our range
uncertainty analysis.

Finally, Monte Carlo simulations typically report dose-to-tissue as compared to dose-to-

water reported by a planning system based on an analytical dose calculation method. In
order to compare the two systems one needs to convert the dose. Such a conversion can be
done based on relative stopping powers and additional terms considering nuclear
interactions (Paganetti, 2009). Reporting dose-to-tissue can result in a slightly different
range compared to reporting dose-to-water, specifically in bone (Paganetti, 2009). The
discrepancy can be on the order of 1–2mm in an extreme case. We do not consider this an
additional uncertainty because clinical decisions are solely made based on dose-to-water.

3.4 Monte Carlo uncertainties due to CT resolution

In a relatively homogeneous geometry the size of the voxels used to divide the geometry for
NIH-PA Author Manuscript

dose calculation plays a minor role. Even for typical head and neck cases significant
uncertainties might not be expected based on the CT resolution typically used for treatment
planning. Problems could be encountered in highly heterogeneous geometries where the
heterogeneities are small, like in lung. Here, materials with low density, i.e. air, and
materials with considerably higher density, i.e. soft tissue, might be averaged to obtain the
HU for a given voxel. To study this effect a swine lung was CT scanned with fine resolution
of 0.4×0.4×0.2 mm3 (Espana and Paganetti, 2011). Monte Carlo simulations were used to
compare the dose distributions predicted with the voxel size typically used in treatment
planning with those expected to be delivered using a finer resolution. CT images with
0.8×0.8×1.2 mm3 as well as 0.8×0.8×2.4 mm3 voxel size were created from the fine
resolution scan by combining and interpolating HU. Figure 8 shows the CT scan of the
swine lung. An artificial rectangular tumor was simulated to study range variations in a
homogenous field with a horizontal dose-fall off downstream of the sharp edge of the
artificial tumor.

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 7

The range when defined as the mean distal falloff position at 90% of the maximum dose
varied by 2.4mm between the fine and the course resolution (used for planning). At the 50%
falloff position the differences were up to 5.6mm. The maximum difference was 38 mm
NIH-PA Author Manuscript

when airways were present in the beam path. Depending on the region of the lung, the
standard deviation was up to 1cm. The differences are due to the poor axial resolution of CT
images typically used in clinical routine. The lung case can be seen as a worst-case scenario.
The maximum differences hold for one ray, which would be smoothed out due to
intrafractional motion and interfractional variations. For sites outside of the lung, the range
uncertainty due to CT resolution might be estimated to only about 0.3% for CT resolutions
typically used in treatment planning.

3.5 Range uncertainties due to biological effects

Proton therapy uses a generic value for the relative biological effectiveness (RBE) to relate
prescribed proton doses to photon doses (Paganetti et al., 2002). The RBE is a function of
dose, tissue endpoint and energy deposition characteristics. The latter is important with
respect to range uncertainties. For a given dose and biological endpoint in proton therapy
one can parameterize the energy deposition characteristics by the linear energy transfer
(LET). The increasing LET with depth in an SOBP is compensated by a decreasing proton
fluence resulting in a homogeneous depth-dose distribution. The increase in LET causes an
increase in the RBE at the distal end of the SOBP. This in turn causes a shift in the
biologically effective range by ~1–2mm due to an LET increase and a dose decrease
NIH-PA Author Manuscript

(Paganetti and Goitein, 2000; Robertson et al., 1975; Wouters et al., 1996). The effect is
illustrated in figure 9. Unfortunately, even if we can predict LET distributions to a
reasonable accuracy using Monte Carlo simulations (Grassberger and Paganetti, 2011), the
knowledge of biological parameters on dose and endpoint dependency is not sufficient to
perform biological, RBE based, treatment planning with sufficient accuracy.

A potential approximation is the use of voxel-by-voxel LET information for biologically

optimized treatment planning, i.e. use elevated LET values as constraints in treatment
planning so that they can be steered away from critical structures as much as possible
(Grassberger et al., 2011). Figure 10 shows two treatment plans and the respective dose-
averaged LET distributions illustrating the concept. This method does not directly address
biological range uncertainties but minimizes their impact for a given treatment plan in
intensity-modulated proton therapy.

The shift of the distal fall-off due to an RBE in excess of 1.1 has been seen for tissues with
low α/β ratio. For tissues with high α/β ratio the RBE at the distal falloff might not exceed
1.1 (it might even be below 1.1 in the center of the SOBP) causing no significant range shift.
Thus, the RBE effect should not be included in the clinical range uncertainty margin because
NIH-PA Author Manuscript

it will presumably always result in a potential range increase. In contrast, the biological
uncertainty (~0.8%) should be included in a margin increase for organs at risk if beams are
pointing towards them.

3.6 In vivo range verification

Uncertainties could be better understood if in vivo range verification could be done with
high precision. The most frequently used method uses positron emission tomography (PET)
imaging, i.e. imaging of the activation of the patient’s tissue by the proton beam (Parodi and
Enghardt, 2000; Parodi et al., 2007; Knopf et al., 2009; Knopf et al., 2011). Proton nuclear
interactions in the patient create positron emitters such as 11C (T1/2=20.39 min), 13N
(T1/2=9.97 min), and 15O (T1/2=2.04 min). A similar method in terms of the underlying
physics is the detection of prompt gamma radiation emitted after nuclear excitation by the
proton beam in tissue (Polf et al., 2009a; Polf et al., 2009b). Interestingly, for the latter

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 8

method, the maximum in the cross section as a function of energy is at lower energies
compared to the PET isotope production cross sections. This causes the prompt gamma
signal maximum to be closer to the Bragg peak than the PET signal maximum (Moteabbed
NIH-PA Author Manuscript

et al., 2011). Another advantage compared to the PET method is a potentially higher count
rate. However, currently detector efficiencies are not able to utilize this advantage leaving
PET as the only practical method at this time. Finally, it has been shown that magnetic
resonance imaging (MRI) done weeks or months after treatment can reveal range
information due to tissue changes, e.g. fatty replacements at the spinal cord (Gensheimer et
al., 2010). Detectors placed inside the human body have also been suggested (Lu, 2008a, b).

PET range verification has been done at several institutions. Relevant for this article is the
capability of Monte Carlo methods to predict the correct PET isotope production. Proton
beams loose energy mainly via electromagnetic interactions. Consequently, the activation
image from PET imaging is not directly correlated to the dose distribution. The established
method is to use a Monte Carlo calculated distribution of the positron emitters and compare
this predicted image with a measured image. The accuracy depends on the underlying cross
section data (Litzenberg et al., 1999; Oelfke et al., 1996; Parodi et al., 2005). The
implementation of inelastic nuclear interactions into a Monte Carlo system can be complex.
One has to rely on a model (or a combination of models) with empirical parameter sets or
experimental cross section data associated with considerable uncertainties. The reason for
the uncertainties in cross sections is based on the fact that the relevant physics experiments
NIH-PA Author Manuscript

(mostly done in the 50’s, 60’s and 70’s) have focused on thin targets thus determining cross
sections for a distinct energy. However, in proton therapy we are interested in a thick target
stopping the entire beam and thus in a continuous cross section information over the entire
energy region of interest.

To study the uncertainty in the predicted range deduced from PET imaging due to
uncertainties in measured cross section data used in the Monte Carlo simulation, a study was
done using a heterogeneous phantom consisting of various materials (Espana et al., 2011).
Studied were the reactions 16O(p,pn)15O, 12C(p,pn)11C and 16O(p,3p3n)11C. Several
published experimental and theoretical cross section data were implemented in a Monte
Carlo system. The predicted PET images were then compared with measured PET images.
The different cross section data resulted in range differences below 1 mm when a 5 min scan
time after treatment was employed but up to 5 mm for a 30 min scan time with 15 min
additional delay (a typical scenario for off-line out-of-the room PET range verification).
Figure 11 shows one example using a gel with high oxygen concentration and two different
experimental cross section data sets from the EXFOR database (EXFOR/CSISRS, 2010), the
cross section published by the ICRU (ICRU, 2000) and a combination of the ICRU cross
section recommendation with the experimental data that was obtained by using an envelope
NIH-PA Author Manuscript

encompassing the available data points (Espana et al., 2011). The conclusion of this study
was that more experimental data in particular for the 16O(p,3p3n)11C reaction channel are
needed to reduce the uncertainties in Monte Carlo simulations to below 1 mm. Note that in
addition to uncertainties in cross sections when comparing measured PET images with
Monte Carlo simulations, there are other inherent limitations of PET range verification, e.g.
scanner resolution, biological washout etc. (Knopf et al., 2009).

4. Discussion
In proton therapy treatment planning the prescribed range is increased to account for
uncertainties in the predicted proton beam range. The guidelines depend on the proton
therapy facility and generally include a generic margin recipe that might however be revised
for specific treatment scenarios. The recipe at Massachusetts General Hospital, 3.5% of the
prescribed range + 1mm, was originally introduced at the Harvard Cyclotron Laboratory

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 9

based on estimations published in a paper by Goitein (Goitein, 1985) and was supposed to
include two factors2. The first one was an absolute value of about 1 mm based on the
uncertainty in patient setup, detection of the skin surface and compensator design. The
NIH-PA Author Manuscript

second factor was based on the uncertainty associated to CT imaging and conversion. With
the CT technology and resolution of the early 1980’s, the uncertainty in determining HU
values was estimated as 2%. Another 1% was added for the CT conversion uncertainty
leading to water equivalent densities in tissue. This lead to √(2%2+1%2)×1.5 standard
deviations or ~3.5%. Thus, interestingly, the definition of this original range uncertainty
recipe did not include uncertainties due to dose calculation, other than those associated with
the water equivalent density. On the other hand, CT scanners have improved quite a lot since
the early 1980’s. Based on the findings summarized above we can try to estimate the current
range uncertainties in proton therapy and the potential impact of Monte Carlo.

First there are those uncertainties that are independent of dose calculation. These are mainly
uncertainties in commissioning, compensator design, beam reproducibility and patient setup.
They have been estimated in table 1. Note that the patient setup error is the range uncertainty
due to a setup uncertainty of ~1mm, not the setup uncertainty itself.

The uncertainties due to dose calculation are caused by defining the patient geometry using
CT and by the dose calculation algorithm itself. The uncertainty from CT imaging itself and
scanner calibration might be ±0.5% for today’s technology (see 3.3). Another contribution is
NIH-PA Author Manuscript

caused by the conversion of HU into relative stopping power or material composition

(leading to stopping power). Based on the discussion above one might assume ±0.5% for
analytical calculations and ±0.2% for Monte Carlo simulations (see 3.3). The CT grid
certainly plays a role but this effect is presumably only significant (bigger than a few %) in
lung and, potentially, in bone or at interfaces of bone and soft tissue. We might assume a
±0.3% effect for a site with typical density variations (see 3.4).

If we assume that pristine peaks in water have been normalized to measured data both for
the Monte Carlo and the analytical planning system, the residual range difference between
Monte Carlo and analytical dose calculation method is mainly caused by a) uncertainties in
stopping power in tissue (i.e. mostly due to the I-value) and b) multiple Coulomb scattering
causing range degradation. The latter can be divided into two effects. First, homogeneities
lateral to the beam path causing scatter disequilibrium (see 3.1) are not modeled correctly in
a typical analytical model, which can lead to significant differences between the two dose
calculation methods, i.e. ±2.5% (see figures 3 and 5). This effect would however affect only
a small volume within the treatment field and should not be considered as a range
uncertainty in all treatment scenarios as it would only appear for certain geometries and
beam arrangements. Thus, it might not be included in a generic margin recipe. Second, the
NIH-PA Author Manuscript

depth of inhomogeneities not considered in a typical analytical model based on water

equivalent pathlengths. This effect might be smaller causing an uncertainty on the order of
0.7% (see 3.1). In contrast, the physics uncertainties in a Monte Carlo model are presumably
below ±0.1%.

The uncertainties in stopping power are mainly due to the I-value but also du to the HU
conversion method. Note that the uncertainty in stopping power ratio discussed above (1.1–
1.8% (Schaffner and Pedroni, 1998)) includes both. One may assume a range uncertainty in
water of ~1.2% (see 3.2) caused by the I-value alone and thus an uncertainty in tissue of
maybe ±1.5%, which affects all dose calculation methods equally.

2M. Goitein, personal communication

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 10

Furthermore, there are uncertainties due to biological considerations (see 3.5). Uncertainties
due to variations in RBE at the distal falloff depend on the biological endpoint (e.g. the α/β
ratio of the tissue). On average, the uncertainty might be ~0.8%. An exception could be
NIH-PA Author Manuscript

tissues with very small α/β ratios where this uncertainty might be bigger. In the calculation
of the added range uncertainty, biological effects were excluded because they would only
increase the range.

Table 1 lists the main sources of uncertainty for nonmoving targets. Most uncertainties do
depend on the beam energy or range so that it is appropriate to give uncertainties in % rather
than mm. The estimations are for soft tissues and bone. As discussed above, lung might
show a much bigger impact on CT imaging uncertainties, possibly up to 5%, and
uncertainties due to the large water-equivalent range in low-density lung tissue. Also not
included are effects from CT artifacts and organ motion. Note also that not all sources in
table 1 are truly independent. For instance, a setup error impacting the beam path could
influence range degradation effects significantly. The appropriate range uncertainty margin
might depend on the ability to avoid dosimetrically challenging beam directions.

The total estimated uncertainties are displayed in figure 12. It seems that the current practice
at most centers might be too conservative. A range uncertainty of 2.7%+1.2mm was
deduced based on the estimations given above. It appears that introducing routine Monte
Carlo dose calculation would not have much of an impact for a typical case as long as the
NIH-PA Author Manuscript

main source of uncertainty is remains, i.e. the uncertainties in the mean excitation energies.
A highly significant impact of Monte Carlo dose calculation can be expected though in
complex geometries where local range uncertainties due to multiple Coulomb scattering will
reduce the accuracy of pencil beam algorithms significantly. The magnitude of the
discrepancy might of course also depend on the quality of the analytical algorithm. It might
well be possible to design an analytical algorithm that comes close to the Monte Carlo
results.

The impact of differences in the predicted range between Monte Carlo dose calculation and
analytical dose calculation and the resulting range that has to be prescribed in treatment
planning to account for dose calculation uncertainties, depends on the treatment site. It
seems advisable that range uncertainties included in treatment planning should not be based
on generic recipes but depend on patient geometry and treatment planning considerations.
Potentially, a method based on a heterogeneity index (Pflugfelder et al., 2007), which
parameterizes the heterogeneous geometry along the beam path, could be applied to define
the most appropriate distal (range) margin due to expected dose calculation errors for a
given site and beam direction or identify those cases where Monte Carlo dose calculation is
desired. Obviously, the uncertainty also depends on the specific geometry in the beam path
NIH-PA Author Manuscript

and the treatment plan, i.e. the beam angles. Robust optimization techniques can be applied
to minimize range uncertainties (Unkelbach et al., 2007).

Acknowledgments
This work was funded in part by NIH/NCI PO1 CA21239.

References
Agostinelli S, Allison J, Amako K, Apostolakis J, Araujo H, Arce P, Asai M, Axen D, Banerjee S,
Barrand G, Behner F, Bellagamba L, Boudreau J, Broglia L, Brunengo A, Burkhardt H, Chauvie S,
Chuma J, Chytracek R, Cooperman G, Cosmo G, Degtyarenko P, Dell’Acqua A, Depaola G,
Dietrich D, Enami R, Feliciello A, Ferguson C, Fesefeldt H, Folger G, Foppiano F, Forti A, Garelli
S, Giani S, Giannitrapani R, Gibin D, Gomez Cadenas JJ, Gonzalez I, Gracia Abril G, Greeniaus G,
Greiner W, Grichine V, Grossheim A, Guatelli S, Gumplinger P, Hamatsu R, Hashimoto K, Hasui

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 11

H, Heikkinen A, Howard A, Ivanchenko V, Johnson A, Jones FW, Kallenbach J, Kanaya N,

Kawabata M, Kawabata Y, Kawaguti M, Kelner S, Kent P, Kimura A, Kodama T, Kokoulin R,
Kossov M, Kurashige H, Lamanna E, Lampen T, Lara V, Lefebure V, Lei F, Liendl M, Lockman
NIH-PA Author Manuscript

W, Longo F, Magni S, Maire M, Medernach E, Minamimoto K, Mora de Freitas P, Morita Y,

Murakami K, Nagamatu M, Nartallo R, Nieminen P, Nishimura T, Ohtsubo K, Okamura M,
O’Neale S, Oohata Y, Paech K, Perl J, Pfeiffer A, Pia MG, Ranjard F, Rybin A, Sadilov S, Di Salvo
E, Santin G, Sasaki T, Savvas N, Sawada Y, Scherer S, Sei S, Sirotenko V, Smith D, Starkov N,
Stoecker H, Sulkimo J, Takahata M, Tanaka S, Tcherniaev E, Safai Tehrani E, Tropeano M,
Truscott P, Uno H, Urban L, Urban P, Verderi M, Walkden A, Wander W, Weber H, Wellisch JP,
Wenaus T, Williams DC, Wright D, Yamada T, Yoshida H, Zschiesche D. GEANT4 - a simulation
toolkit. Nuclear Instruments and Methods in Physics Research. 2003; A 506:250–303.
Ahlen SP. Theoretical and experimental aspects of the energy loss of relativistic heavily ionizing
particles. Reviews of Modern Physics. 1980; 52:121–173.
Andreo P. On the clinical spatial resolution achievable with protons and heavier charged particle
radiotherapy beams. Phys Med Biol. 2009; 54:N205–N215. [PubMed: 19436099]
Andreo P, Medin J, Bielajew AF. Constraints of the multiple-scattering theory of Moliere in Monte
Carlo simulations of the transport of charged particles. Medical Physics. 1993; 20:1315–1325.
[PubMed: 8289712]
Bednarz B, Daartz J, Paganetti H. Dosimetric accuracy of planning and delivering small proton
therapy fields. Phys Med Biol. 2010; 55:7425–7438. [PubMed: 21098920]
Bednarz B, Lu HM, Engelsman M, Paganetti H. Uncertainties and correction methods when modeling
passive scattering proton therapy treatment heads with Monte Carlo. Phys Med Biol. 2011;
NIH-PA Author Manuscript

56:2837–2854. [PubMed: 21478569]

Bichsel H, Hiraoka T. Energy loss of 70 MeV protons in elements. Nuclear Instruments and Methods
in Physics Research. 1992; B66:345–351.
Chvetsov AV, Paige SL. The influence of CT image noise on proton range calculation in radiotherapy
planning. Phys Med Biol. 2010; 55:N141–N149. [PubMed: 20182006]
du Plessis FCP, Willemse CA, Loetter MG, Goedhals L. The indirect use of CT numbers to establish
material properties needed for Monte Carlo calculation of dose distributions in patients. Medical
Physics. 1998; 25:1195–1201. [PubMed: 9682205]
Espana S, Paganetti H. The impact of uncertainties in the CT conversion algorithm when predicting
proton beam ranges in patients from dose and PET-activity distributions. Physics in Medicine and
Biology. 2010; 55:7557–7572. [PubMed: 21098912]
Espana S, Paganetti H. Uncertainties in planned dose due to the limited voxel size of the planning CT
when treating lung tumors with proton therapy. Phys Med Biol. 2011; 56:3843–3856. [PubMed:
21628773]
Espana S, Zhu X, Daartz J, El Fakhri G, Bortfeld T, Paganetti H. The reliability of proton-nuclear
interaction cross-section data to predict proton-induced PET images in proton therapy. Phys Med
Biol. 2011; 56:2687–2698. [PubMed: 21464534]
EXFOR/CSISRS. Experimental Nuclear Reaction Data. International Atomic Energy Agency—
NIH-PA Author Manuscript

Nuclear Data Section (Vienna, Austria). 2010 (http://www-nds.iaea.org/exfor/exfor.htm).

Ferrari A, Sala PR, Fasso A, Ranft J. FLUKA: a multi-particle transport code. CERN Yellow Report
CERN 2005-10; INFN/TC 05/11, SLAC-R-773 (Geneva: CERN). 2005
Gensheimer MF, Yock TI, Liebsch NJ, Sharp GC, Paganetti H, Madan N, Grant PE, Bortfeld T. In
vivo proton beam range verification using spine MRI changes. Int J Radiat Oncol Biol Phys. 2010;
78:268–275. [PubMed: 20472369]
Goitein M. The measurement of tissue heterodensity to guide charged particle radiotherapy. Int J
Radiat Oncol Biol Phys. 1977; 3:27–33. [PubMed: 96060]
Goitein M. Calculation of the uncertainty in the dose delivered during radiation therapy. Med Phys.
1985; 12:608–612. [PubMed: 4046996]
Goitein M, Sisterson JM. The Influence of Thick Inhomogeneities on Charged Particle Beams.
Radiation Research. 1978; 74:217–230. [PubMed: 96480]
Gottschalk B, Koehler AM, Schneider RJ, Sisterson JM, Wagner MS. Multiple Coulomb scattering of
160 MeV protons. Nuclear Instruments and Methods in Physics Research. 1993; B74:467–490.

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 12

Grassberger C, Paganetti H. Elevated LET components in clinical proton beams. Physics in Medicine
and Biology. 2011; 56:6677–6691. [PubMed: 21965268]
Grassberger C, Trofimov A, Lomax A, Paganetti H. Variations in linear energy transfer within clinical
NIH-PA Author Manuscript

proton therapy fields and the potential for biological treatment planning. International Journal of
Radiation Oncology, Biology, Physics. 2011; 80:1559–1566.
Herault J, Iborra N, Serrano B, Chauvel P. Monte Carlo simulation of a protontherapy platform
devoted to ocular melanoma. Med Phys. 2005; 32:910–919. [PubMed: 15895573]
ICRU. Bethesda, MD: International Commission on Radiation Units and Measurements; 1989. Tissue
Substitutes in Radiation Dosimetry and Measurement. Report No. 44
ICRU. Bethesda, MD: International Commision on Radiation Units and Measurements; 1992. Photon,
Electron, Proton and Neutron Interaction Data for Body Tissues. Report No. 46
ICRU. Bethesda, MD: International Commission on Radiation Units and Measurements; 1993.
Stopping Powers and Ranges for Protons and Alpha Particles. Report No. 49
ICRU. Bethesda, MD: International Commission on Radiation Units and Measurements; 2000.
Nuclear Data for Neutron and Proton Radiotherapy and for Radiation Protection. Report No. 63
Jaekel O, Reiss P. The influence of metal artefacts on the range of ion beams. Physics in Medicine and
Biology. 2007; 52:635–644. [PubMed: 17228110]
Janni JF. Proton range energy tables, 1 keV - 10 GeV. Atomic Data and Nuclear Data Tables. 1982;
27:147–529.
Jiang H, Paganetti H. Adaptation of GEANT4 to Monte Carlo dose calculations based on CT data.
Medical Physics. 2004; 31:2811–2818. [PubMed: 15543788]
NIH-PA Author Manuscript

Kimstrand P, Tilly N, Ahnesjo A, Traneus E. Experimental test of Monte Carlo proton transport at
grazing incidence in GEANT4, FLUKA and MCNPX. Phys Med Biol. 2008; 53:1115–1129.
[PubMed: 18263962]
Knopf A, Parodi K, Bortfeld T, Shih HA, Paganetti H. Systematic analysis of biological and physical
limitations of proton beam range verification with offline PET/CT scans. Phys Med Biol. 2009;
54:4477–4495. [PubMed: 19556685]
Knopf AC, Parodi K, Paganetti H, Bortfeld T, Daartz J, Engelsman M, Liebsch N, Shih H. Accuracy
of proton beam range verification using post-treatment positron emission tomography/computed
tomography as function of treatment site. Int J Radiat Oncol Biol Phys. 2011; 79:297–304.
[PubMed: 20646839]
Kumazaki Y, Akagi T, Yanou T, Suga D, Hishikawa Y, Teshima T. Determination of the mean
excitation energy of water from proton beam ranges. Radiation Measurements. 2007; 42:1683–
1691.
Litzenberg DW, Roberts DA, Lee MY, Pham K, Vander Molen AM, Ronningen R, Becchetti FD. On-
line monitoring of radiotherapy beams: experimental results with proton beams. Med Phys. 1999;
26:992–1006. [PubMed: 10436901]
Lu HM. A point dose method for in vivo range verification in proton therapy. Phys Med Biol. 2008a;
53:N415–N422. [PubMed: 18997263]
NIH-PA Author Manuscript

Lu HM. A potential method for in vivo range verification in proton therapy treatment. Phys Med Biol.
2008b; 53:1413–1424. [PubMed: 18296770]
Matsufuji N, Tomura H, Futami Y, Yamashita H, Higashi A, Minohara S, Endo M, Kanai T.
Relationship between CT number and electron density, scatter angle and nuclear reaction for
hadron-therapy treatment planning. Physics in Medicine and Biology. 1998; 43:3261–3275.
[PubMed: 9832015]
Moteabbed M, Espana S, Paganetti H. Monte Carlo patient study on the comparison of prompt gamma
and PET imaging for range verification in proton therapy. Physics in Medicine and Biology. 2011;
56:1063–1082. [PubMed: 21263174]
Mustafa AAM, Jackson DF. The relation between x-ray CT numbers and charged particle stopping
powers and its significance for radiotherapy treatment planning. Physics in Medicine and Biology.
1983; 28:169–176. [PubMed: 6408654]
Newhauser WD, Giebeler A, Langen KM, Mirkovic D, Mohan R. Can megavoltage computed
tomography reduce proton range uncertainties in treatment plans for patients with large metal
implants? Phys Med Biol. 2008; 53:2327–2344. [PubMed: 18421122]

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 13

Oelfke U, Lam GKY, Atkins MS. Proton dose monitoring with PET: quantitative studies in Lucite.
Physics in Medicine and Biology. 1996; 41:177–196. [PubMed: 8685254]
Paganetti H. Calculation of the spatial variation of relative biological effectiveness in a therapeutic
NIH-PA Author Manuscript

proton field for eye treatment. Physics in Medicine and Biology. 1998; 43:2147–2157. [PubMed:
9725595]
Paganetti H. Dose to water versus dose to medium in proton beam therapy. Physics in Medicine and
Biology. 2009; 54:4399–4421. [PubMed: 19550004]
Paganetti H, Goitein M. Radiobiological significance of beam line dependent proton energy
distributions in a spread-out Bragg peak. Medical Physics. 2000; 27:1119–1126. [PubMed:
10841418]
Paganetti H, Jiang H, Lee S-Y, Kooy H. Accurate Monte Carlo for nozzle design, commissioning, and
quality assurance in proton therapy. Medical Physics. 2004; 31:2107–2118. [PubMed: 15305464]
Paganetti H, Jiang H, Parodi K, Slopsema R, Engelsman M. Clinical implementation of full Monte
Carlo dose calculation in proton beam therapy. Phys Med Biol. 2008; 53:4825–4853. [PubMed:
18701772]
Paganetti H, Kooy H. Proton radiation in the management of localized cancer. Expert Rev Med
Devices. 2010; 7:275–285. [PubMed: 20214431]
Paganetti H, Niemierko A, Ancukiewicz M, Gerweck LE, Loeffler JS, Goitein M, Suit HD. Relative
biological effectiveness (RBE) values for proton beam therapy. International Journal of Radiation
Oncology, Biology, Physics. 2002; 53:407–421.
Parodi K, Enghardt W. Potential application of PET in quality assurance of proton therapy. Phys Med
NIH-PA Author Manuscript

Biol. 2000; 45:N151–N156. [PubMed: 11098922]

Parodi K, Paganetti H, Shih HA, Michaud S, Loeffler JS, DeLaney TF, Liebsch NJ, Munzenrider JE,
Fischman AJ, Knopf A, Bortfeld T. Patient study of in vivo verification of beam delivery and
range, using positron emission tomography and computed tomography imaging after proton
therapy. Int J Radiat Oncol Biol Phys. 2007; 68:920–934. [PubMed: 17544003]
Parodi K, Poenisch F, Enghardt W. Experimental Study on the Feasibility of In-Beam PET for
Accurater Monitoring of Proton Therapy. IEEE Transactions in Nuclear Science. 2005; 1
Pelowitz DBE. MCNPX User's Manual, Version 2.5.0. Los Alamos National Laboratory. 2005 LA-
CP-05-0369.
Perl J, Schuemann J, Shin J, Faddegon B, Paganetti H. Topas - Fast and easy to use Monte Carlo
platform for proton therapy applications. Medical Physics. 2011; 38:3754.
Petti PL. Differential-pencil-beam dose calculations for charged particles. Medical Physics. 1992;
19:137–149. [PubMed: 1320182]
Petti PL. Evaluation of a pencil-beam dose calculation technique for charged particle radiotherapy.
International Journal of Radiation Oncology, Biology, Physics. 1996; 35:1049–1057.
Pflugfelder D, Wilkens JJ, Szymanowski H, Oelfke U. Quantifying lateral tissue heterogeneities in
hadron therapy. Med Phys. 2007; 34:1506–1513. [PubMed: 17500481]
Pia MG, Begalli M, Lechner A, Quintieri L, Saracco P. Physics-related epistemic uncertainties in
NIH-PA Author Manuscript

proton depth dose simulation. IEEE Transactions on Nuclear Science. 2010; 57

Polf JC, Peterson S, Ciangaru G, Gillin M, Beddar S. Prompt gamma-ray emission from biological
tissues during proton irradiation: a preliminary study. Phys Med Biol. 2009a; 54:731–743.
[PubMed: 19131673]
Polf JC, Peterson S, McCleskey M, Roeder BT, Spiridon A, Beddar S, Trache L. Measurement and
calculation of characteristic prompt gamma ray spectra emitted during proton irradiation. Phys
Med Biol. 2009b; 54:N519–N527. [PubMed: 19864704]
Robertson JB, Williams JR, Schmidt RA, Little JB, Flynn DF, Suit HD. Radiobiological Studies of a
High-Energy Modulated Proton Beam Utilizing Cultured Mammalian Cells. Cancer. 1975;
35:1664–1677. [PubMed: 807318]
Sawakuchi GO, Titt U, Mirkovic D, Mohan R. Density heterogeneities and the influence of multiple
Coulomb and nuclear scatterings on the Bragg peak distal edge of proton therapy beams. Phys
Med Biol. 2008; 53:4605–4619. [PubMed: 18678928]

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 14

Schaffner B, Pedroni E. The precision of proton range calculations in proton radiotherapy treatment
planning: experimental verification of the relation between CT-HU and proton stopping power.
Physics in Medicine and Biology. 1998; 43:1579–1592. [PubMed: 9651027]
NIH-PA Author Manuscript

Schaffner B, Pedroni E, Lomax A. Dose calculation models for proton treatment planning using a
dynamic beam delivery system: an attempt to include density heterogeeity effects in the analytical
dose calculation. Physics in Medicine and Biology. 1999; 44:27–41. [PubMed: 10071873]
Schneider U, Pedroni E, Lomax A. The calibration of CT Hounsfield units for radiotherapy treatment
planing. Physics in Medicine and Biology. 1996; 41:111–124. [PubMed: 8685250]
Schneider W, Bortfeld T, Schlegel W. Correlation between CT numbers and tissue parameters needed
for Monte Carlo simulations of clinical dose distributions. Physics in Medicine and Biology. 2000;
45:459–478. [PubMed: 10701515]
Soukup M, Alber M. Influence of dose engine accuracy on the optimum dose distribution in intensity-
modulated proton therapy treatment plans. Physics in Medicine and Biology. 2007; 52:725–740.
[PubMed: 17228117]
Soukup M, Fippel M, Alber M. A pencil beam algorithm for intensity modulated proton therapy
derived from Monte Carlo simulations. Physics in Medicine and Biology. 2005; 50:5089–5104.
[PubMed: 16237243]
Stankovskiy A, Kerhoas-Cavata S, Ferrand R, Nauraye C, Demarzi L. Monte Carlo modelling of the
treatment line of the Proton Therapy Center in Orsay. Phys Med Biol. 2009; 54:2377–2394.
[PubMed: 19321923]
Szymanowski H, Oelfke U. Two-dimensional pencil beam scaling: an improved proton dose algorithm
NIH-PA Author Manuscript

for heterogeneous media. Physics in Medicine and Biology. 2002; 47:3313–3330. [PubMed:
12375823]
Titt U, Sahoo N, Ding X, Zheng Y, Newhauser WD, Zhu XR, Polf JC, Gillin MT, Mohan R.
Assessment of the accuracy of an MCNPX-based Monte Carlo simulation model for predicting
three-dimensional absorbed dose distributions. Phys Med Biol. 2008a; 53:4455–4470. [PubMed:
18670050]
Titt U, Zheng Y, Vassiliev ON, Newhauser WD. Monte Carlo investigation of collimator scatter of
protontherapy beams produced using the passive scattering method. Phys Med Biol. 2008b;
53:487–504. [PubMed: 18185001]
Tourovsky A, Lomax AJ, Schneider U, Pedroni E. Monte Carlo dose calculations for spot scanned
proton therapy. Physics in Medicine and Biology. 2005; 50:971–981. [PubMed: 15798269]
Unkelbach J, Chan TC, Bortfeld T. Accounting for range uncertainties in the optimization of intensity
modulated proton therapy. Phys Med Biol. 2007; 52:2755–2773. [PubMed: 17473350]
Urban L. Multiple scattering model in Geant4. CERN report. 2002 CERN-OPEN-2002-070.
Urie M, Goitein M, HolleyW R, Chen GTY. Degradation of the Bragg peak due to inhomogeneities.
Physics in Medicine and Biology. 1986; 31:1–15. [PubMed: 3952143]
Urie M, Goitein M, Wagner M. Compensating for heterogeneities in proton radiation therapy. Phys
Med Biol. 1984; 29:553–566. [PubMed: 6330772]
NIH-PA Author Manuscript

van Luijk P, van t' Veld AA, Zelle HD, Schippers JM. Collimator scatter and 2D dosimetry in small
proton beams. Phys Med Biol. 2001; 46:653–670. [PubMed: 11277215]
Vanderstraeten B, Reynaert N, Paelinck L, Madani I, De Wagter C, De Gersem W, De Neve W,
Thierens H. Accuracy of patient dose calculation for lung IMRT: A comparison of Monte Carlo,
convolution/superposition, and pencil beam computations. Med Phys. 2006; 33:3149–3158.
[PubMed: 17022207]
Wilson RR. Radiological Use of Fast Protons. Radiology. 1946; 47:487–491. [PubMed: 20274616]
Wouters BG, Lam GKY, Oelfke U, Gardey K, Durand RE, Skarsgard LD. RBE Measurement on the
70 MeV Proton Beam at TRIUMF using V79 cells and the High Precision Cell Sorter Assay.
Radiation Research. 1996; 146:159–170. [PubMed: 8693066]
Yang M, Virshup G, Clayton J, Zhu XR, Mohan R, Dong L. Theoretical variance analysis of single-
and dual-energy computed tomography methods for calculating proton stopping power ratios of
biological tissues. Phys Med Biol. 2010; 55:1343–1362. [PubMed: 20145291]
Zacharatou Jarlskog C, Paganetti H. Physics settings for using the Geant4 toolkit in proton therapy.
IEEE Transactions in Nuclear Science. 2008; 55:1018–1025.

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 15
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 1.
Range of a proton beam in water based on the continuous slowing down approximation as a
function of proton energy.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 16
NIH-PA Author Manuscript

Figure 2.
Left: Distal falloff of a Bragg peak from a 220 MeV proton beam after penetrating different
inhomogeneities shown on the right. The inlet shows the doses without normalization to the
maximum. Adapted from (Sawakuchi et al., 2008), with permission.
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 17
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 3.
Axial views of dose distributions calculated using a commercial planning system based on a
pencil-beam algorithm (XiO (Computerized Medical Systems); left) and a Monte Carlo
system (right). The red arrow indicates the range difference due to a density interface
parallel to the beam path. Adapted from (Paganetti et al., 2008), with permission.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 18
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 4.
Measured (circles) and Monte Carlo simulated (line) lateral dose profile. The experimental
setup shown on the left consisted of a beam impinging on a half-block of bone equivalent
material (black) into a water phantom (grey). The dashed line indicates the position of the
measured profile. Adapted from (Paganetti et al., 2008), with permission.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 19
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 5.
Monte Carlo (MC) and analytically (TPS) generated dose distributions in a small field
delivered to a head and neck patient. (a): depth dose distributions along the beam central
axis; (b): transverse dose profiles; (c): axes used for figures a) and b). From (Bednarz et al.,
2010), with permission.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 20
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 6.
Depth dose curves in water for a 122 MeV proton beam assuming mean excitation energies
of 67 eV, 75 eV, and 80 eV, respectively. From (Andreo, 2009), with permission.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 21
NIH-PA Author Manuscript

Figure 7.
NIH-PA Author Manuscript

Monte Carlo model of one of the treatment heads at the Francis H Burr Proton Therapy
Center at Massachusetts General Hospital. The simulation was done with the TOPAS Monte
Carlo system based on the Geant4 Monte Carlo code (Perl et al., 2011).
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 22
NIH-PA Author Manuscript

Figure 8.
Left: Image of an inflated swine lung. The CT image was edited to introduce a 4 cm thick
water-equivalent entrance wall and a 2×2×2 cm3 tumor. A 5 mm CTV expansion is shown
in red. Right: Subsection of the CT image showing the transvere and coronal view of the
structures for different voxel sizes. Adapted from (Espana and Paganetti, 2011), with
permission.
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 23
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 9.
Left: Physical dose and biological dose (physical dose times RBE; line with circles)
simulated for one particular endpoint and dose. Right: SOBP and the underlying proton
energy distributions at four different depths causing an increasing LET with depth. Adapted
from (Paganetti and Goitein, 2000) and (Paganetti, 1998), with permission.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 24
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 10.
Two dose distributions (left; in %) and the corresponding dose averaged LET distributions,
NIH-PA Author Manuscript

LETd, (right; in keV/µm) illustrating that clinically equivalent dose distributions can be
achieved with quite different LET distributions steering dose falloffs away from critical
structures in intensity modulated proton therapy. Based on (Grassberger et al., 2011).

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 25
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 11.
Measured and Monte Carlo simulated depth activity profiles (activity concentration in Bq/
ml) obtained for a gel consisting of 11.03% H, 1.04% C, 0.32% N, and 87.6% O. The scan
was taken immediately after irradiation with a scan time of 5 minutes. Adapted from
(Espana et al., 2011), with permission.
NIH-PA Author Manuscript

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 26
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Figure 12.
Dotted lines: Typically applied range uncertainty margins in proton therapy treatment
planning as currently typically applied at the Massachusetts General Hospital (3.5% +
1mm), the MD Anderson Proton Therapy Center in Houston (3.5% + 3mm), the Loma Linda
University Medical Center (3.5% + 3mm), the Roberts Proton Therapy Center at the
University of Pennsylvania (3.5% + 3mm), and the University of Florida Proton Therapy
Institute (2.5% + 1.5mm). Note that these centers may apply bigger margins in specific
NIH-PA Author Manuscript

treatment scenarios. Dashed line: estimated uncertainty without the use of Monte Carlo dose
calculation. Solid line: estimated uncertainty for complex geometries without the use of
Monte Carlo dose calculation. Dashed-dotted line: estimated uncertainty with the use of
Monte Carlo dose calculation.

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.
 Paganetti Page 27

Table 1
Estimated proton range uncertainties and their sources and the potential of Monte Carlo for reducing the
NIH-PA Author Manuscript

uncertainty.

Source of range uncertainty in the patient Range Range

uncertainty uncertainty
without Monte with Monte
Carlo Carlo
Independent of dose calculation:

Measurement uncertainty in water for commissioning ± 0.3 mm ± 0.3 mm

Compensator design ± 0.2 mm ± 0.2 mm

Beam reproducibility ± 0.2 mm ± 0.2 mm

Patient setup ± 0.7 mm ± 0.7 mm

Dose calculation:

Biology (always positive) ^ + ~0.8% g + ~0.8% g

CT imaging and calibration ± 0.5% b ± 0.5% b

CT conversion to tissue (excluding I-values) ± 0.5% a ± 0.2% d

CT grid size ± 0.3% ± 0.3%

NIH-PA Author Manuscript

Mean excitation energy (I-values) in tissues ± 1.5% c ± 1.5% c

Range degradation; complex inhomogeneities − 0.7% e ± 0.1 %

Range degradation; local lateral inhomogeneities * ± 2.5% f ± 0.1 %

Total (excluding *, ^) 2.7% + 1.2 mm 2.4% + 1.2 mm

Total (excluding ^) 4.6% + 1.2 mm 2.4% + 1.2 mm

The number are estimations based on finding by a(Schaffner and Pedroni, 1998; Matsufuji et al., 1998);
b
(Chvetsov and Paige, 2010);
c
(ICRU, 1993; Bichsel and Hiraoka, 1992; Kumazaki et al., 2007);
d
(Espana and Paganetti, 2010);
e
(Sawakuchi et al., 2008; Bednarz et al., 2010; Urie et al., 1986);
f
(Bednarz et al., 2010);
g
(Paganetti and Goitein, 2000; Robertson et al., 1975; Wouters et al., 1996).
NIH-PA Author Manuscript

The estimations are average numbers based on 1.5 standard deviations. Extreme cases, like lung, might show bigger uncertainties.

Phys Med Biol. Author manuscript; available in PMC 2013 June 07.