A Monte Carlo study of IMRT beamlets in inhomogeneous media

a
Andrew O. Jones
b)
Radiation Sciences, Drexel University College of Medicine, Philadelphia, Pennsylvania 19102
Indra J. Das
Department of Radiation Oncology, University of Pennsylvania, Philadelphia, Pennsylvania 19104
Frederick L. Jones, Jr.
Department of Thoracic Medicine, Geisinger Medical Center, Danville, Pennsylvania 17821
Received 19 March 2002; accepted for publication 16 November 2002; published 5 February 2003
Intensity Modulated Radiation Therapy IMRT has trended toward smaller multiple radiation elds
thereby increasing the resolution of the intensity map. Vendors have introduced multileaf collima-
tors with beam apertures of 0.5 cm and less. The beam characteristics of these smaller elds have
not been adequately assessed, especially in the presence of inhomogeneities. Most dosimetric de-
vices have signicant limitations due to nite size, dose rate, and energy dependence. We studied
the effect of inhomogeneities on small beamlets. The 6, 15, and 24 MV beams were modeled using
the EGSnrc Monte Carlo code. Point source beams of circular eld sizes 0.5, 1.0, 3.0, 5.0, and 10 cm
were simulated in a water phantom at 100 SSD. A 3 cm inhomogeneity of lung tissue was incor-
porated between 3 and 6 cm in the phantom. The depth dose curves and proles were compared by
beam size and density of the inhomogeneity. The Monte Carlo simulations show that for small
elds a marked dose decrease in the presence of low-density media due to the lack of lateral
electronic equilibrium is observed. As the density and eld size increase, the dose reduction is less
pronounced and for the 10 cm eld there is an increased dose as expected due to lack of attenuation.
This data suggests that current TPS may dramatically over- or underestimate the dose in inhomo-
geneous media for small eld sizes that are used for IMRT. 2003 American Association of
Physicists in Medicine. DOI: 10.1118/1.1539040
Key words: inhomogeneity corrections, Monte Carlo calculations
I. INTRODUCTION
Historically, radiation oncologists have disregarded the ef-
fects of different densities in the human body.
14
Since
Compton interactions, which dominate in therapeutic energy
beams, are dependent on electron density and not physical
density, this approximation has been generally correct. Algo-
rithms attempting to correct for density differences Equiva-
lent Pathlength, TAR ratio, Power Law have been based on
scaling beam data measured in a water phantom according to
physical densities. These corrections ignore electron trans-
port.
Currently there is a great interest in combining multiple
small beamlets to create a predened dose distribution.
Intensity-modulated radiation therapy IMRT uses a series
of beamlets varying in intensity to paint a dose picture or
intensity map. Beamlets of greater intensity create darker
pixels higher dose than those of lower intensity. By de-
creasing the size of the beamlets, a higher resolution inten-
sity map can be created, resulting in a ner ability to tailor
the dose distribution. Multileaf collimators MLCs creating
beams of 0.5 cm and less, including the Millenium 120 from
Varian Oncology Systems, Palo Alto, CA, the Beak slit col-
limator from NOMOS Corp, Sewickley, PA, and the m3
micro-MLC from Brainlab AG Heimstetten, Germany, are
now commercially available.
Although some studies have been conducted on the do-
simetry of small photon beams,
510
the effect of density dif-
ferences on such small beams has not been extensively stud-
ied. Previous reports focused mainly on radiosurgical beams
greater than 1.25 cm in diameter.
24
The effects of air cavi-
ties on radiosurgical beams, showing a dose drop across the
cavity, subsequent buildup, and dose enhancement on the
distal side of the cavity, have been reported.
3,4
Algorithm
comparisons in the presence of lung and air cavities have
also been studied down to 44 cm
2
eld sizes.
1
Small beams
1 cm passing through different densities of lung or bone,
however, have not been well studied.
Small radiation beams are inherently difcult to
measure.
511
Standard ion chambers are large in comparison
to the beam and do not have the spatial resolution needed to
resolve the narrow central region of uniform dose and the
steep dose gradients of the penumbra;
8
they may have a dose
rate dependence.
12
Film, particularly Gaf chromic lm, pre-
pared and read carefully, may be the best available
dosimeter.
1,2,47
It is widely available and does not require
specialized development or reading equipment.
Ideally, one would like to study multiple densities, com-
positions, energies, and eld sizes without disturbing the
beam. This is best done using Monte Carlo computer
simulation.
4,5,8,11
II. MATERIALS AND METHODS
This research utilized the EGSnrc
13
Electron Gamma
Shower, National Research Council Monte Carlo code made
296 296 Med. Phys. 30 3, March 2003 0094-240520033032965$20.00 2003 Am. Assoc. Phys. Med.
available from the National Research Council of Canada.
This is the latest version of the EGS series that simulate
radiation interactions in any media. The user code DOSRZ
14
was used to dene the problem. DOSRZ creates the input les
for EGSnrc as dened in the cylindrical coordinate system.
Photon spectra can be dened, as can geometry. Many media
are provided with EGSnrc, or arbitrary media can be created
using the PEGS4 preprocessor for EGS software.
Predened water, lung density 0.26 gram per cubic cen-
timeter, and bone density 1.85 grams per cubic centimeter
media were used. For lung tissue of different density PEGS4
was used to dene the lung using the elemental composition.
Photon spectra for 6, 15, and 24 MV beams from Varian
linear accelerators from Mohan et al.
15
provided with the
EGSnrc package, were used. The parameters set for EGSnrc
were ECUT0.521 MeV, PCUT0.001 MeV, using the
PRESTA algorithm. The only variance reduction used was
photon forced interaction in the media.
Point source, circular beams were simulated at 100 cm
sourcesurface distance SSD from a semi-innite phantom.
The inhomogeneous media was placed between 3 and 6 cm
in the phantom. The dose was calculated for 0.2 cm thick
slabs from 0 to 12 cm. The radii of the dose volumes were
different for different eld sizes. Simulations were run until
the statistical errors were below 1%.
The verication of the Monte Carlo code was done by
comparing 0.5 cm depth dose curves generated for the 6 MV
photon beams to data measured on a clinical 6 MV photon
beam Varian Oncology Systems, Palo Alto, CA. Figure 1
shows good agreement between the Monte Carlo simulation
and the measured beam data are very close at all depths.
Percentage depth dose curves were created for lung den-
sities of 0.15, 0.20, 0.26, 0.30, 0.35, and 0.40 g/cm
3
, for
bone of density of 1.85 g/cm
3
and for water. Circular diam-
eter eld sizes of 0.5, 1.0, 3.0, 5.0, and 10.0 cm and beam
energies of 6, 15, and 24 MV were simulated. These curves
were compared with the homogeneous phantom data using a
FIG. 1. Depth dose curves for Monte Carlo simulated and measured 6 MV
photon beams in water. The curves show good agreement at all depths.
FIG. 2. Depth-dose curves for 6 dark solid, 15 dashed, and 24 light
solid MV beams using a 0.5 cm eld size through lung and in a homoge-
neous phantom. The large drop in dose across the lung density is evident as
is the dose enhancement beyond the lung.
FIG. 3. Depth dose curves for 24 MV photon beam showing the effects of
eld size. Increasing the eld size decreases the dose drop across the inho-
mogeneity, but the dose drop is still present at the relatively large eld size
of 5 cm.
FIG. 4. 6 MV photon depth dose curves for eld sizes 0.5, 1, 3, and 5 cm
generated using Monte Carlo calculations in a phantom with tissue density
0.26 g/cm
3
between 3 and 6 cm depth. By the 5 cm eld size the dose
decrease in lung has nearly disappeared.
297 Jones, Das, and Jones: A Monte Carlo study of IMRT beamlets 297
Medical Physics, Vol. 30, No. 3, March 2003
Dose Perturbation Factor DPF, dened as the ratio of the
dose at a point in the medium to the dose at that point in
water.
III. RESULTS
Increased photon transmission and the loss of lateral elec-
tronic equilibrium in small elds leads to a signicant dose
drop in the lung tissue. Figure 2 shows a large dose drop in
the lung with a 0.5 cm eld size for 6, 15, and 24 MV photon
beams. The decrease begins proximal to the tissuelung in-
terface as the contribution from backscattered photons and
electrons decreases. Across the lung tissue the depth dose
decreases only slightly before rising abruptly at the distal
lungtissue interface. Beyond the interface the dose is en-
hanced, mainly due to the increased photon uence through
the lung.
For the 24 MV photon beam the dose decrease effect is
evident with eld sizes as large as 5 cm, as seen in Fig. 3.
The higher-energy electrons have greater range, leading to
lateral electronic disequilibrium persisting through larger
eld sizes. Increased photon transmission also leads to a
dose enhancement distal to the second lung tissue interface,
even for the large 10 cm eld. Although the region beyond
the lung benets from an increased dose, the region nearest
the interface is underdosed as the dose builds up due to in-
creased photon attenuation and decreased electron range.
Similarly, Fig. 4 shows a drop in the depth dose across the
lung for the 6 MV photon beam. The buildup beyond the
second interface is similar as well, however, the effects of the
eld size are not as great for the low-energy beam. By the
time the eld size reaches 5 cm, the dose decit in the lung
has been erased and a slight increase is seen.
The effect of the lung tissue on the dose as a comparison
to the homogeneous case can be determined by looking at
the DPF. A gauge of the generalized effects is a plot of the
average DPF through the center of the lung volume, as seen
in Fig. 5. As expected, the average DPF across the lung
increases with decreasing energy due to the increased photon
transmission and electron range of the high-energy beams.
The curves converge at the extremes of eld size. For the
larger elds, lateral electronic equilibrium is eventually es-
tablished even for the high energy beams and the effects of
the low-density lung become more consistent with the lower-
energy beams, which established equilibrium at smaller eld
sizes. For the small elds the DPF converges as the elds
become so small that even the low-energy electrons that
dominate in the 6 MV photon beam escape the eld. This
convergence is to be expected since extrapolating the elds
to a zero eld size where all of the secondary electrons es-
cape the eld would create a convergence, even for the
lowest-energy photon eld.
Figures 6 and 7 show the effects of density on the dose to
the lung. The average DPF taken across the central region of
the lung is plotted against the density of the lung for 6 MV
Fig. 6 and 15 MV Fig. 7 photon beams. As expected,
decreased photon transmission and electron range combine
FIG. 5. Graph of average DPF versus Field size for 6, 15, 24 MV beam
energies showing decreased DPF with increased energy for all eld sizes.
FIG. 6. Average Dose Perturbation Factors DPFs for 6 MV photons as a
function of density for 0.5, 1.3, and 5 cm eld sizes with lung tissue located
between 3 and 6 cm depth showing the dependence of the DPF on eld size
and density. The humps at 0.2 g/cm
3
density may be the result of statistical
noise in the Monte Carlo calculation.
FIG. 7. DPF curves for a 15 MV photon beam showing the effects of density
and eld size. For smaller eld sizes the change in DPF with density is
rapid. Once lateral electronic equilibrium is reached, the density effect is
much less.
298 Jones, Das, and Jones: A Monte Carlo study of IMRT beamlets 298
Medical Physics, Vol. 30, No. 3, March 2003
to increase the DPF as the media becomes denser. With more
photons liberating secondary electrons and the pathlengths of
the electrons becoming shorter, more of the KERMA is ulti-
mately captured within the eld as dose. The increase in DPF
with density is not, however, linear. The DPF increases more
rapidly than the density for all eld sizes at both energies.
The bumps in the 10 cm eld size curves are likely due to
statistical noise in the Monte Carlo data for that particular
eld size. These curves suggest a more complex relationship
in low-density dosimetry than simple radiological scaling of
the physical density.
IV. DISCUSSION
Signicant dose perturbations are observed proximally,
distally, and within different density media. For lower-
density media like lung, all beam energies show a signicant
dose decrease in the inhomogeneity, as seen in Fig. 2. This is
at odds with the traditional dose algorithms, such as the
equivalent pathlength algorithm, which scale the dose ac-
cording to the inverse of the density. Although the dose de-
crease is most prominent at small eld sizes, for higher en-
ergies such as the 24 MV beam shown in Fig. 3, it persists up
to more traditional clinical eld sizes of 10 cm.
The study also shows the dose changing proximal to the
tissuemedia interface, either dropping at the lung interface
or increasing for the bone. The changes, the result of differ-
ences in the electron backscatter at the interface, are not
modeled even in modern dose algorithms like convolution
superposition.
The dose decrease in the lung is nonlinearly related to
eld size, as evidenced in the DPF curves in Fig. 5. The
point at which the DPF reaches 1.0, i.e., no dose change
from water, occurs when the eld radius is about 2 cm for a
6 MV beam. At this point lateral electronic equilibrium is
reached and electrons that are created within the eld will
deposit their energy within the eld. The higher-energy 15
and 24 MV beams show a dose decrease through correspond-
ingly larger eld sizes, nally reaching a DPF of 1.0 at close
to a 10 cm eld size. This observation points out that the
density scaling aw of traditional algorithms is present for a
high-energy beam even at more traditional clinical eld
sizes.
The DPF increases with increasing density of the lung,
but again nonlinearly, as seen in Figs. 6 and 7. Larger eld
sizes show less variation due to density. The 15 MV 0.5 cm
eld shows a greater than 40% increase in DPF from a den-
sity of 0.15 g/cm
3
to 0.40 g/cm
3
, whereas the 5 cm eld
increases only about 15%. The increases are not linear with
density reecting a need to use a different scaling factor than
physical density.
Dose beyond the lung is increased regardless of eld size
or energy. Even in cases where the lung dose is enhanced,
the dose beyond the lung is also enhanced. This is due to
decreased photon attenuation and increased transmission of
forward scattered electrons through the lung depositing their
dose distal to the interface. A secondary buildup region is
seen at the lungtissue interface that corresponds to this hy-
pothesis. Traditional experience with density-scaled algo-
rithms also supports dose enhancement beyond the low-
density media.
For small elds of diameter less than the range of the
electrons set into motion, electrons and scattered photon gen-
erated anywhere within the eld can potentially escape. For
elds that are approximately the width of the range, the ef-
fects are small since the photon interactions are dominated
by Compton interactions and the scatter is forward-peaked.
As the eld size decreases, photons and electrons scattered at
smaller angles will leave the eld and not be replaced. The
total KleinNishina cross section per electron is independent
of the atomic number because the electron binding energy is
assumed to be 0, since the photon energy is much greater
than the electron binding energy.
15
Hence, the Compton mass
attenuation coefcient is linearly dependent on the electron
density. Electron density is approximately independent of
atomic number except for hydrogen; Z/A is about constant
except at high Z so the total Compton mass attenuation for
all elements and mixtures found naturally in the body is
about equal. Thus, the largest effect on the cross sections will
be the electron density, which is proportional, generally, to
the density of the material.
The KleinNishina interaction cross section and the scat-
tering angle cross section decrease with increasing energy.
Thus, fewer photons are interacting and the recoil angle of
the electron is decreasing.
16
At the same time the range of the
Compton electrons is increased. The stopping power of the
medium for electrons is also dependent upon the electron
density and inversely dependent on their velocity. Therefore,
lower-density media and higher-energy electrons combine to
increase the range.
Taking all these issues into consideration shows that small
elds in low-density media will experience decreased photon
attenuation and increased electron range. Increasing the
beam energy in this situation would decrease the photon at-
tenuation, increase the electron range, and decrease the elec-
tron recoil angle. For the energies commonly used in radio-
therapy 425 MV the change in the angular distribution of
the recoil electrons is slight. The result is fewer photons
interacting, electrons depositing their dose farther down-
stream, and more electrons leaving the radiation eld; lateral
electronic disequilibrium. Within the low-density region
there is a dose decit.
The magnitude of the change in dose is not linearly re-
lated to the physical density despite the relationship of the
attenuation coefcients and the mass stopping power to
physical density. Although the stopping power for the elec-
trons and the attenuation coefcients for the photons are re-
lated to the density they are interdependent and the combi-
nation is nonlinear. This observation may be due to changes
in the electron and photon spectra or changes in the scatter-
ing characteristics of the photons and electrons. Electron
stopping power varies as the natural logarithm of the mean
excitation potential of the media, however, converting the
dose in lung to dose in water changes the results by less than
2%,
17
ruling out the effect of different elemental composition
299 Jones, Das, and Jones: A Monte Carlo study of IMRT beamlets 299
Medical Physics, Vol. 30, No. 3, March 2003
of the lung. The net effect is a function of beam energy,
physical density, and eld size.
Although this study looked only at the effects of the cen-
tral axis dose, the low-density regions will affect the beam
prole as well. Several studies looked at the inuence of air
cavities on radiosurgical beams.
3,4,9
These studies have
shown that the penumbra of small beams is attened out,
becoming wider on the low dose area and narrower toward
the center of the beam with the change in full-width at tenth
maximum increasing up to a factor of 1.23 with decreasing
eld size.
4
The total penumbra width measured from the
90%10% lines has been shown to increase with increasing
eld size and air gap depth.
3
The full-width at half-maximum
was nearly the same in both the heterogeneous and homoge-
neous phantoms for all the studies. This result correlates with
the lateral loss of electrons from the center of the eld to the
edges and the corresponding drop on the central axis dose
seen in this study. How penumbral widening changes relative
to density was not studied, but it can be surmised that the
effect would mimic the central axis depth loses and become
less pronounced with increasing density as the electron range
decreases. Further study in this area is warranted.
V. CONCLUSIONS
Signicant changes occur in and near tissue inhomogene-
ities. These changes are not easily modeled using only physi-
cal density. They are dependent upon multiple interconnected
factors including density, eld size, and beam energy. Dose
algorithms that attempt to scale dose only by physical den-
sity do not accurately predict dose within or near inhomoge-
neities. This is especially true with the smaller eld sizes
being used for IMRT. More study is being conducted to
evaluate the magnitude of the differences between the Monte
Carlo results and current IMRT algorithms.
a
Presented in part at the 2001 AAPM meeting in Salt Lake City, Utah.
b
Please address correspondence to Andrew Jones, MS, Department of Ra-
diation Oncology, Bryn Mawr Hospital, 130 South Bryn Mawr Avenue,
Bryn Mawr, Pennsylvania 19010. Telephone: 610 526-3372; electronic
mail: JonesAO@netscape.net
1
M. R. Arneld, C. H. Siantar, J. Siebers, P. Garmon, L. Cox, and R.
Mohan, The impact of electron transport on the accuracy of computed
dose, Med. Phys. 27, 12661274 2000.
2
S. N. Rustgi, A. K. Rustgi, S. B. Jiang, and K. M. Ayyangar, Dose
perturbation caused by high-density inhomogeneities in small beams in
stereotactic radiosurgery, Phys. Med. Biol. 43, 35003518 1998.
3
A. K. Rustgi, M. A. Samuels, and S. N. Rustgi, Inuence of air inho-
mogeneities in radiosurgical beams, Med. Dosim 22, 95100 1997.
4
T. D. Solberg, F. E. Holly, A. A. F. DeSalles, R. E. Wallace, and J. B.
Smathers, Implications of tissue heterogeneity for radiosurgery in head
and neck tumors, Int. J. Radiat. Oncol., Biol., Phys. 32, 235239
1995.
5
B. E. Bjamgard, J.-S. Tsai, and R. K. Rice, Doses on the central axes of
narrow 6-MV x-ray beams, Med. Phys. 17, 794799 1990.
6
R. K. Rice, J. L. Hansen, G. K. Svensson, and R. L. Siddon, Measure-
ments of dose distributions in small beams of 6 MV x-rays, Phys. Med.
Biol. 32, 10871099 1987.
7
S. N. Rustgi and D. M. D. Frye, Dosimetric characterization of radio-
surgical beams with a diamond detector, Med. Phys. 22, 21172121
1995.
8
M. Westermark, J. Arndt, B. Nilsson, and A. Brahme, Comparative do-
simetry in narrow high-energy beams, Phys. Med. Biol. 45, 685702
2000.
9
H. Saitoh, T. Fujisaki, R. Sakai, and E. Kunieda, Dose distribution of
narrow beam irradiation for small lung tumor, Int. J. Radiat. Oncol.,
Biol., Phys. 53, 13801387 2002.
10
C. Martens, C. De Wagter, and W. De Neve, The value of the PinPoint
ion chamber for characterization of small eld segments used in
intensity-modulated radiotherapy, Phys. Med. Biol. 45, 25192530
2000.
11
F. Verhaegen, I. J. Das, and H. Palmans, Monte Carlo dosimetry study
of a 6 MV stereotactic radiosurgery unit, Phys. Med. Biol. 43, 2755
2768 1998.
12
W. U. Laub, T. W. Kaulich, and F. Nusslin, Energy and dose rate de-
pendence of a diamond detector in the dosimetry of 425 MV photon
beams, Med. Phys. 24, 535536 1997.
13
I. Kawrakow and D. W. O. Rogers, The EGSnrc code system: Monte
Carlo simulation of electron and photon transport, NRCC Report PIRS-
701, Ottawa, ON, Ionizing Radiation Standards, National Research
Council of Canada.
14
D. W. O. Rogers, I. Kawradow, J. P. Seuntjens, and B. R. B Walters,
NRC user codes for EGSnrc incomplete draft, NRCC Report PIRS-
702, Ottawa, ON, Ionizing Radiation Standards, National Research
Council of Canada.
15
R. Mohan, C. Chui, and L. Lidofsky, Energy and angular distributions
of photons from medical linear accelerators, Med. Phys. 12, 592597
1985.
16
F. Attix, Introduction Radiological Physics and Radiation Dosimetry
Wiley, New York, 1986.
17
J. Siebers, P. Keall, A. Nahum, and R. Mohan, Converting absorbed
dose to medium to absorbed dose to water for Monte Carlo based photon
beam dose calculations, Phys. Med. Biol. 45, 983995 2000.
300 Jones, Das, and Jones: A Monte Carlo study of IMRT beamlets 300
Medical Physics, Vol. 30, No. 3, March 2003