A flattening filter free photon treatment concept evaluation

with Monte Carlo

U. Titt,a兲 O. N. Vassiliev, F. Pönisch, L. Dong, H. Liu, and R. Mohan
M. D. Anderson Cancer Center, The University of Texas, 1515 Holcombe Boulevard, Houston,
Texas 77030
共Received 9 February 2005; revised 28 March 2006; accepted for publication 28 March 2006;
published 10 May 2006兲
In principle, the concept of flat initial radiation-dose distribution across the beam is unnecessary for
intensity modulated radiation therapy. Dynamic leaf positioning during irradiation could appropri-
ately adjust the fluence distribution of an unflattened beam that is peaked in the center and deliver
the desired uniform or nonuniform dose distribution. Removing the flattening filter could lead to
reduced treatment time through higher dose rates and reduced scatter, because there would be
substantially less material in the beam; and possibly other dosimetric and clinical advantages. This
work aims to evaluate the properties of a flattening filter free clinical accelerator and to investigate
its possible advantages in clinical intensity modulated radiation therapy applications by simulating
a Varian 2100-based treatment delivery system with Monte Carlo techniques. Several depth-dose
curves and lateral dose distribution profiles have been created for various field sizes, with and
without the flattening filter. Data computed with this model were used to evaluate the overall quality
of such a system in terms of changes in dose rate, photon and electron fluence, and reduction in
out-of-field stray dose from the scattered components and were compared to the corresponding data
for a standard treatment head with a flattening filter. The results of the simulations of the flattening
filter free system show that a substantial increase in dose rate can be achieved, which would reduce
the beam on time and decrease the out-of-field dose for patients due to reduced head-leakage dose.
Also close to the treatment field edge, a significant improvement in out-of-field dose could be
observed for small fields, which can be attributed to the change in the photon spectra, when the
flattening filter is removed from the beamline. © 2006 American Association of Physicists in
Medicine. 关DOI: 10.1118/1.2198327兴

Key words: Monte Carlo simulations, flattening filter free, treatment head modification, beam line
modification

I. INTRODUCTION
Conventional clinical accelerators used in radiotherapy are
equipped with a flattening filter 共FF兲. The FF is designed to
produce approximately uniform dose distribution across the
field at a certain depth in a homogeneous phantom. Such a
field is typically not the best solution for complex treatment
problems. For this reason, additional beam-modifying de-
vices are used, such as wedges, compensators, and dynamic
multileaf collimators 共MLCs兲. MLCs are quite flexible in
terms of shaping fluence distribution across the field. In prin-
ciple, they render the FF unnecessary.
The FF introduces a very substantial amount of material
into the beamline. It is one of the most significant sources of
undesired radiation scatter.1,2 More than 50% of primary ra-
diation can be scattered or absorbed in the FF. Relevant nu-
merical data will be presented later in this study.
By removing the FF an increase in dose rate can be
achieved, leading to a shorter irradiation time per treatment
fraction. This will also reduce treatment head scatter, out-of-
field exposure of the patient, and leakage dose through the
MLC. The implications of removing the FF with respect to
increased fluence and dose rates; changes in the out-of-field
dose close to the penumbra; and changes in the lateral dose
profiles, as a function of depth in a water phantom, are in-
vestigated in this work.
Several previous investigations of FF free machines have
been undertaken, primarily to investigate the physical prop-
erties of the beam, such as changes in dmax, the location of
maximum dose deposition and energy spectrum changes.3–5
Variation of the beam quality with off-axis distance have
been researched by Zefkili et al.6 and Lee et al.7 O’Brien
et al.8 removed the FF from a Therac-6 linear accelerator
共AECL, Atomic Energy of Canada Ltd., Mississauga, On-
tario, Canada兲 in an effort to reduce the treatment time of a
stereotactic radiosurgery procedure. Gillies et al.9 discussed
engineering modifications to a Therac-6 linear accelerator to
perform dynamic stereotactic radiotherapy, including a head
modification, which allowed easy removal of the FF. Sixel
and Faddegon4 calculated photon spectra for small fields on
the Therac-6 linear accelerator as well as depth-dose and
profiles with and without the FF. Further research was con-
ducted to investigate portal imaging properties,10 dose deliv-
ery enhancement in high-Z material-loaded tumors,11 and an
application in helical tomotherapy.12 Fu et al.13 investigated
intensity modulated radiation therapy 共IMRT兲 delivery time
with and without the FF. They used a BJ-6B linear accelera-

1595 Med. Phys. 33 „6…, June 2006 0094-2405/2006/33„6…/1595/8/$23.00 © 2006 Am. Assoc. Phys. Med. 1595
1596 Titt et al.: FFF Monte Carlo simulation 1596

tor 共in 6-MV photon mode兲 and the Focus 共version 3.2.1,
CMS, Computerized Medical Systems, St. Louis, Missouri,
USA兲 treatment-planning system.
In this paper, we present our first results on the way to a
flattening filter free 共FFF兲 clinical accelerator, based on the
Varian Clinac 2100 共Varian Medical Systems Inc., Palo Alto,
California, USA兲. Numerous FFF irradiations have been
simulated to investigate the properties of the device and the
first results are being presented. The study focuses on the
6-MV photon mode.

II. MATERIALS AND METHODS

The Monte Carlo radiation transport programs used in this
study are BEAMnrc14 and MCNPX 2.5d.15 Both programs
require input files, defining the geometry and various physics FIG. 1. Central-axis depth-dose profile for benchmarking the MCNPX pro-
gram. Relative dose D is shown as a function of depth, d, in the water
parameters. The geometry used in both programs was based phantom. The number of calculated initial electron histories is 108, field size
on the Varian Clinac 2100 treatment head. Although the ge- is 10⫻ 10 cm2, initial mean electron energy is 6.2 MeV, and energy spread
ometry in BEAMnrc was restricted to the major beamline is 3% 共FWHM兲. The statistical uncertainty of the simulation was smaller
than 1% at dmax, and the difference to data measured during the commis-
components, a more detailed model was developed for
sioning of a Varian Clinac 2100C showed a maximum of 1.1%. The depth,
MCNPX, including the structural components and the lead dmax, of the maximum dose differed by 1.2± 0.1 mm between measurement
shielding, in order to investigate realistic out-of-field doses. and simulation.
Both systems use table-based cross-sectional data. The elec-
tron transport was terminated when the kinetic energy of the
electron reached a value below 100 keV 共MCNPX兲 or cal fluctuations in the Monte Carlo results, the calculated
200 keV 共BEAMnrc兲. Photon transport was terminated at depth-dose curves were normalized to the sum of the dose
10 keV in both programs. Initially, the complete radiation value between depths of 5 cm and 20 cm, using the follow-
transport was modeled, starting with electrons stopping in a ing equation:
target to produce bremsstrahlung photons. BEAMnrc and
20 cm
MCNPX both provide a variety of variance reduction tech-
niques, which have been utilized in our simulations. In the D= 兺
5 cm
D共z兲dz, 共1兲
case of BEAMnrc, selective bremsstrahlung splitting was
used, and in the case of MCNPX, a weight window mesh and compared with measurements. Both results, measured
was created to improve the statistics in the receptors and thus and calculated, could then be compared with respect to the
lower the statistical uncertainties. The bremsstrahlung pho- relative value of the maximum dose Dmax and the corre-
tons were transported through the first collimator, a thin sponding depth dmax. Dose distributions from field sizes of
vacuum window, the FF followed by an ionization chamber,
and through a set of jaws into a cubic water phantom 共50
⫻ 50⫻ 50 cm3 in size兲 where all dose profiles were tallied.
The properties of the electron source were determined
based on the work of Cho et al.16 It was assumed that the
primary electron beam is parallel, and the energy spectrum
and lateral spread of the electron fluence are of Gaussian
shape. The energy distribution is centered at 6.2 MeV and
has a 3% full width at half-maximum 共FWHM兲. The lateral
spread is 1 mm FWHM.
For subsequent calculations, phase-space files were used,
to eliminate the CPU-intensive simulation of bremsstrahlung
production. The Varian accelerators at M. D. Anderson Can-
cer Center deliver photons with different energy distribu-
tions, denoted as 6 MV and 18 MV. All data presented in the
following refer to the 6 MV mode. Extensive benchmarking
simulations have been undertaken to compare simulations of FIG. 2. Central-axis depth-dose profile for benchmarking the BEAMnrc pro-
the model including the FF with measurements performed gram, showing relative dose, D, as a function of depth, d, in water. The
during commissioning of a photon treatment machine at M. number of calculated initial electron histories is 2.7⫻ 108, field size is 10
⫻ 10 cm2, initial mean electron energy is 6.2 MeV, and energy spread is 3%
D. Anderson. The measured central axis depth-dose curves
共FWHM兲. The statistical uncertainty of these simulations was smaller than
were normalized to unity at the depth, dmax, of the maximum 0.43% at dmax, and the maximum relative dose difference to measured data
dose deposition, Dmax. To minimize the influence of statisti- was 1.1%. The location of dmax differed by 0.03 mm.

Medical Physics, Vol. 33, No. 6, June 2006

1597 Titt et al.: FFF Monte Carlo simulation 1597

TABLE I. Comparison of calculated 共MCNPX兲 and measured central-axis

depth-dose profiles at various field sizes. In all cases, the initial electron
energy was 6.2 MeV. A denotes the irradiation field size, dmax denotes the
location of the maximum dose, and ⌬Dmax is the relative dose discrepancy
between the measurement and the calculations at dmax.

A / cm2 dsim
max / cm dmeas
max / cm ⌬Dmax

4⫻4 1.55 1.64 0.70%

10⫻ 10 1.44 1.56 1.13%
20⫻ 20 1.39 1.36 2.42%
40⫻ 40 1.26 1.26 3.60%

4 ⫻ 4 cm2, 10⫻ 10 cm2, 20⫻ 20 cm2, and 40⫻ 40 cm2 have

been investigated.
The measured lateral dose profiles were normalized to FIG. 3. Benchmark comparison of lateral dose profiles, calculated with MC-
NPX and BEAMnrc at a depth of 6.0 cm in a water phantom using a field
unity on the central axis, and the calculations were normal- size of 10⫻ 10 cm2 with a measured dose profile. The maximum relative
ized to the sum of the dose values between the two 50% difference, ⌬max, in the central high dose region of the profile, excluding the
locations in the penumbra, −x50 and +x50, penumbra and beyond, shows values between 1.2% and 1.8%. The penum-
bra widths between the 80% level and the 20% level are denoted by P80–20.
x50
D= 兺 D共x兲dx,
−x50
共2兲
ence, electron filters of various materials were simulated up-
and compared to the measured values. The size of the pen- stream of the ionization chamber, and their influence on
umbra in the 90% to 10% interval and in the 80% to 20% photon and electron fluences was evaluated.
interval, the differences of x50 from measurements and simu- The statistical uncertainty of the simulations 共one ␴ inter-
lations, as well as the maximum differences at the center of val兲 was below 1% for the depth-dose profiles of all field
the beam and at the field edges were evaluated for three field sizes calculated with BEAMnrc and for field sizes of up to
sizes, 10⫻ 10 cm2, 20⫻ 20 cm2, and 40⫻ 40 cm2. 20⫻ 20 cm2 calculated with MCNPX. The 40⫻ 40 cm2 field
Removing the FF from the model enabled us to investi- size resulted in a statistical uncertainty of up to 3% when
gate the properties of an FFF machine with respect to MCNPX was used. The simulations were performed until the
changes in the central axis depth-dose profiles, the lateral statistical uncertainties reached values lower than 1% at and
dose profiles at various depths in the water phantom, and around the central beam axis.
increases in the electron and photon fluences and the respec-
tive energy fluences at the downstream surface of the moni- III. RESULTS
tor chamber and at the surface of the water phantom.
A. Benchmark calculations
The electron and photon fluences at the monitor chamber
and at the phantom were of particular interest. Increased Central axis depth-dose profiles of various field sizes with
electron fluence in the ionization chamber could result in the FF in the beamline were computed with MCNPX and
saturation of the detector and, hence, lead to nonlinearities in with BEAMnrc to benchmark the Monte Carlo system. Fig-
the ratio of the delivered therapeutic dose to monitor units ure 1 shows an example of a central-axis depth-dose curve
共MU兲 measured. The electron fluence at the surface of the
phantom also had to be minimized, in order to avoid exces-
sive dose in the build-up region. To reduce the electron flu- TABLE III. Comparison of measured and calculated lateral dose profiles. The
field size is denoted as A, ⌬x50 is the lateral difference measured at the 50%
dose point in the penumbra, and P90–10 as well as P80–20 describe the width
of the penumbra measured from the 90% point to the 10% point and from
TABLE II. Comparison of calculated 共BEAMnrc兲 and measured central-axis the 80% to 20%, respectively. The maximum relative differences are given
depth-dose profiles at various field sizes. Initial electron energy was again by max共⌬D兲.
6.2 MeV. A denotes the irradiation field size, dmax denotes the location of
the maximum dose, and ⌬Dmax is the relative dose discrepancy between the Method A / cm2 ⌬x50 / mm P90–10 / mm P80–20 / mm max共⌬D兲
measurement and the calculations at dmax.
Measurement 10⫻ 10 11 5.5
A / cm2 dsim dmeas ⌬Dmax MCNPX 10⫻ 10 0.5 11 3.5 1.2%
max / cm max / cm
BEAMnrc 10⫻ 10 0.1 8.5 3.4 1.1%
Measurement 20⫻ 20 12 6
4⫻4 1.50 1.64 0.30% MCNPX 20⫻ 20 0.1 7 3.5 2.0%
10⫻ 10 1.50 1.56 1.10% BEAMnrc 20⫻ 20 0.1 10 3.3 1.0%
20⫻ 20 1.37 1.36 0.04% Measurement 40⫻ 40 11
40⫻ 40 1.50 1.29 1.60% BEAMnrc 40⫻ 40 0.2 7 6.0%

Medical Physics, Vol. 33, No. 6, June 2006

1598 Titt et al.: FFF Monte Carlo simulation
FIG. 4. Comparison of simulated central-axis depth-dose 共per initial par-
ticle兲 profiles in a 40⫻ 40 cm2 field, with and without the FF, normalized to
100 at the shoulder of the FFF dose falloff. Besides the overall dose in-
crease, the profile from the filter-free simulation clearly shows a sharp spike
in the build-up region, which is attributed to excessive electron fluence.
calculated with MCNPX using a field size of 10⫻ 10 cm2,
and Fig. 2 shows the same setup calculated with BEAMnrc.
Tables I and II list the results of the comparison of the simu-
lations with measured data.
Lateral beam profiles were also simulated for various field
sizes at numerous depths. Figure 3 shows the comparison of
Monte Carlo calculations to measured data for a field size of
10⫻ 10 cm2. The differences between the measurement and
the simulations in lateral field size at the 50% dose level,
⌬x50, were determined to be 0.5 mm or less for MCNPX
calculations, whereas ⌬x50 was about 0.1 mm in the BEAM-
nrc calculations. The penumbra widths, P90−10 and P80−20,
calculated from the 90% level to the 10% level and from
80% to 20%, respectively, showed a systematic bias: most of
the simulations resulted in smaller penumbras compared to
the measured ones, whereas the penumbra widths from
MCNPX and BEAMnrc calculations only differed by a frac-
tion of a millimeter. Table III summarizes the results of the
comparisons.
B. Simulations without the flattening filter, fluence
evaluation
Initial simulations of irradiations without the FF showed
an overall increase of the dose on the central axis 共per inci-
TABLE IV. Photon fluences ␾␥ per initial electron at the downstream surface
of the ionization chamber 共IC兲 and at the upstream surface of the water
phantom 共WP兲 for two different field sizes A as well as with and without the
FF in the beam. All values at the WP are averaged fluence over the total
surface of the water phantom 共50⫻ 50 cm2兲, and the IC values represent the
averaged fluence through the entire surface of the ionization chamber win-
dow. The average statistical uncertainty of the calculation was smaller than
3%.
A ␾␥ at IC 共FF兲 ␾␥ at IC 共FFF兲 ␾␥ at WP 共FF兲 ␾␥ at WP 共FFF兲
共cm2兲 共cm−2 n−1兲 共cm−2 n−1兲 共cm−2 n−1兲 共cm−2 n−1兲
10⫻ 10 6.02⫻ 10−5 7.82⫻ 10−5 8.75⫻ 10−8 1.93⫻ 10−7
40⫻ 40 6.02⫻ 10−5 7.82⫻ 10−5 1.33⫻ 10−6 1.99⫻ 10−6
Medical Physics, Vol. 33, No. 6, June 2006
1598
TABLE V. Electron fluences ␾e per initial electron at the downstream surface
of the ionization chamber 共IC兲 and at the upstream surface of the water
phantom 共WP兲 for two different field sizes A as well as with and without the
FF in the beam. All values at the WP are averaged fluence over the total
surface of the water phantom 共50⫻ 50 cm2兲, and the IC values represent the
fluences averaged over the entire surface of the downstream ionization
chamber window. The average statistical uncertainty of the calculation was
smaller than 3%.
A ␾e at IC 共FF兲 ␾e at IC 共FFF兲 ␾e at WP 共FF兲 ␾e at WP 共FFF兲
共cm2兲 共cm−2 n−1兲 共cm−2 n−1兲 共cm−2 n−1兲 共cm−2 n−1兲
10⫻ 10 4.63⫻ 10−7 2.87⫻ 10−6 8.88⫻ 10−10 3.13⫻ 10−9
40⫻ 40 4.59⫻ 10−7 2.94⫻ 10−6 4.81⫻ 10−9 1.93⫻ 10−8
dent particle兲 by a factor of 2 at least. The ratio of the flat-
tening filter-free simulation to the results with the FF 共field
size 10⫻ 10 cm2兲 showed two different ranges; in both, the
ratio is almost linearly dependent on the depth z. First, a
steep decline in the build up region could be observed 共from
a ratio of 3.2 at z = 0.05 cm to 2.4 at z = 1.2 cm兲 followed by
a shallow curve 共from 2.4 at z = 1.5 cm to 2.2 at z = 20 cm
and to 2.1 at z = 40 cm兲. The depth of maximum dose dmax
was found at 1.36 cm compared to 1.44 cm when calculated
with a FF. Both these results indicate that irradiations with-
out the flattening filter have an increased low-energy compo-
nent. The ratios of central-axis depth-dose with and without
the FF at the field size of 40⫻ 40 cm2 showed the same
shape. The first steep decline showed a ratio of 3.7 at z
= 0.05 cm and 2.4 at z = 1 cm; the second, a shallow curve
starting at 1 cm showed the ratios of 2.1 at z = 20 cm and 2.0
at z = 40 cm. From the depth-dose profile simulations of
larger field sizes 共40⫻ 40 cm2兲, an additional contribution to
the surface dose could be clearly identified showing a spike
in the build-up region 共Fig. 4兲. To identify the source of the
spike and to eliminate it, the photon fluences and the electron
fluences were further investigated.
FIG. 5. Electron fluences, ␾e, per initial particle, tallied at the upstream
surface of the water phantom, as a function of electron energy Ee. The field
size is 40⫻ 40 cm2. The solid line shows the fluence calculated with a
flattening filter free beamline. Introduction of a 2-mm copper foil 共dotted
line兲 reduced the total electron fluence to 22% above the total fluence with
the FF in the beamline 共dashed-dotted line兲. The nylon foil reduced ␾e to
46% above this curve.
1599 Titt et al.: FFF Monte Carlo simulation
FIG. 6. Photon fluences ␾␥, per initial electron, at the upstream surface of
the water phantom as a function of photon energy E␥. The field size for these
simulations was 40⫻ 40 cm2. The solid line shows the fluence calculated
with a flattening filter free beamline. Introduction of a 2 mm copper foil
共dotted line兲 reduced the total photon fluence by 9%, and results with the
nylon foil showed a reduction of 7%.
The electron fluence, ␾e, and the photon fluence, ␾␥, per
initial particle, n, in units of cm−2 n−1 were tallied and aver-
aged over the whole surface of the monitor chamber window,
and the total area of the upstream surface of the water phan-
tom 共50⫻ 50 cm2兲 for field sizes 10⫻ 10 cm2 and 40
⫻ 40 cm2. Tables IV and V list all electron and photon flu-
ences. A notable increase in the electron fluence on the ion-
ization chamber window is observed when the FF is re-
moved. Many of these electrons reach the phantom surface
and contribute to the above-mentioned dose spike in the
build-up region. We describe this phenomenon in the Discus-
sion section.
To investigate possible effects on the ionization chamber
performance, ␾e was tallied according to the particle direc-
tion, so the component of backscattered electrons entering
the ionization chamber could be identified. The amount of
backscattered electron fluence in the monitor chamber was
calculated to be 3.4% 共±0.2% 兲 with the FF and 3.1%
共±0.2% 兲 without the FF, for the 10⫻ 10 cm2 field. For the
40⫻ 40 cm2 field, the values were lower, 1.7% 共±0.1% 兲 with
flattening filter and 1.2% 共±0.1% 兲 without.
For the FFF case, the relative contribution of electrons,
originating from various beamline components, was evalu-
ated just below the jaws. The majority of the electrons
共74.2%兲 were primary electrons from the linear accelerator,
whereas the materials of the target button contributed 5.4%.
TABLE VI. Fluence comparison of FFF operation using electron stopping foils, consisting of 11 mm of nylon
and 2 mm of copper. All electron and photon fluences are averaged over the complete surface of the ionization
chamber 共IC兲 window and over the total upstream surface of the water phantom 共WP兲. The field size is 40
⫻ 40 cm2. The average statistical uncertainty of the calculations was smaller than 3%.
11 mm Nylon
␾e at IC 共cm−2 n−1兲 6.70⫻ 10−7
␾e at WP 共cm−2 n−1兲 7.15⫻ 10−9
␾␥ at IC 共cm−2 n−1兲 7.47⫻ 10−5
␾␥ at WP 共cm−2 n−1兲 1.87⫻ 10−6
Medical Physics, Vol. 33, No. 6, June 2006
1599
Further contributions were 2.6% from the primary collima-
tor, 1.3% from the vacuum window, and 16.5% from all
remaining beamline components. To prevent these excess
electrons from reaching the phantom surface and saturating
the ionization chamber, two different foils were simulated in
the position of the removed FF. An 11-mm nylon slab and a
2-mm copper foil were tested at a field size of 40⫻ 40 cm2
concerning electron stopping and photon transparency. Fig-
ure 5 shows ␾e as a function of electron energy, Ee, at the
upstream surface of the water phantom, and Fig. 6 shows the
photon fluence, ␾␥, as a function of photon energy E␥ at the
same location. Table VI lists all fluence values from the ny-
lon and the copper foil calculations at both locations, the
ionization chamber, and the water phantom. The copper foil
removes more electrons and low energy photons, including
characteristic x rays, from the beam.
C. Lateral dose profiles FF vs FFF
FFF lateral dose profiles were calculated for several field
sizes using the copper foil as an electron filter. In standard
photon treatment machines, the shape of the lateral profiles
shows a depth dependency. It is particularly strong for fields
larger than 10⫻ 10 cm2. At shallow depths, i.e., at the depth
of maximum dose deposition, dmax, an increased shoulder
can be found, whereas at larger depths, beyond a depth of
10 cm, the shoulder tends to decrease below the dose level at
the central axis. This behavior is mainly due to significant
variation of beam quality across the field. The flattening filter
thickness decreases with increasing off-axis distance. For
this reason the photon spectrum is hardened further on the
central axis relative to the shoulder area. Figure 7 shows an
example of this change in shape by overlaying 20⫻ 20 cm2
field-size profiles. The dose is normalized to the central axis
and the field width x⬘ is normalized to x50. The shapes of
lateral dose profiles calculated without the FF show much
less dependency on the depth, as can be seen in Fig. 8.
Whereas the penumbras of fields from flattening filter free
calculations do not show significant differences when com-
pared to simulations with flattening filter, the out-of-field
dose contribution close to the treatment-field edge shows sig-
nificant differences in small fields. Figure 9 shows the lateral
profiles of a 4 ⫻ 4 cm2 treatment field. It can be seen that the
out-of-field dose from the FFF accelerator is substantially
lower, and it falls off faster with distance. This means that a
2 mm Cu No foil
5.60⫻ 10−7 2.94⫻ 10−6
6.16⫻ 10−9 1.93⫻ 10−8
7.37⫻ 10−5 7.82⫻ 10−5
1.83⫻ 10−6 1.99⫻ 10−6
1600 Titt et al.: FFF Monte Carlo simulation 1600

FIG. 7. Lateral dose profiles calculated for a 20⫻ 20 cm2 field size using the
FIG. 9. Comparison of out-of-field dose calculated with 共solid兲 and without
FF in the beamline. The dose, D, was normalized to the central-axis dose
共dashed兲 the FF for a 4 ⫻ 4 cm2 treatment field. The symbols show the
level, and the lateral width, x⬘, was normalized to x50, the 50% location in
Monte Carlo results, while the lines depict fitted curves to these results.
the penumbra. Note, that with increasing depth in the water phantom the
shape of the profile changes significantly.

estimation in the entrance region was expected. Comparative

significant reduction in the out-of-field dose and enhanced
sparing of critical tissues and organs close to small treatment
fields can be achieved.
D. Total scatter factor
The total scatter factors, Sc,p, defined as the dose rate at a
reference depth for a given field size, divided by the dose
rate at the same depth for the reference field size 共10
⫻ 10 cm2兲17 were calculated for several square fields. Both,
an FFF treatment head and a standard setup with a FF have
been simulated. Table VII lists all values, including the sta-
tistical uncertainties, describing the 1␴ confidence interval.
IV. DISCUSSION
The depth-dose calculations with MCNPX, which were
used for benchmarking this Monte Carlo program, show that
the program overestimates the maximum dose deposition
from 0.7% at a field size of 4 ⫻ 4 cm2 to up to 3.6% at a field
size of 40⫻ 40 cm2 compared to measurements. This over-
calculations of bremsstrahlung production with identical
physics options and geometries showed that MCNPX pro-
duced up to 20% more photons with energies below 200 keV
compared to BEAMnrc, which showed far better agreement
with experimental data. At a depth of 5 cm and more, both
the MCNPX and the BEAMnrc depth-dose calculations
showed excellent agreement with measured data. The data
obtained from BEAMnrc show considerably better agree-
ment at smaller depths, only differences of 0.04% through
1.6% in this region could be observed. The lateral dose pro-
files show very good agreement in the high-dose region close
to the central axis, ±1% to ±2% at field sizes below 20
⫻ 20 cm2, whereas for larger fields 共40⫻ 40 cm2兲, the uncer-
tainties are somewhat larger. At the treatment-field edges, the
difference could be up to 6%. The penumbras 共80% to 20%兲
show reasonable agreement with the measurements, showing
differences from 1.0 mm to 2.7 mm. The measurements were
performed with an ionization chamber, which had a diameter
TABLE VII. Total scatter factors, Sc,p, for square fields, normalized to dmax of
a 10⫻ 10 cm2 field for the FFF system and a standard Varian Clinac 2100
setup with the FF. The statistical uncertainties describe the 1␴ - confidence
interval.

A / cm2 Sc,p 共FFF兲 Sc,p 共FF兲

2⫻2 0.927± 0.48% 0.899± 0.45%

3⫻3 0.950± 0.51% 0.939± 0.53%
4⫻4 0.972± 0.53% 0.951± 0.52%
6⫻6 0.979± 0.59% 0.972± 0.53%
8⫻8 0.982± 0.63% 0.987± 0.61%
10⫻ 10 1 1
12⫻ 12 1.007± 0.63% 1.011± 0.59%
15⫻ 15 1.011± 0.67% 1.023± 0.58%
20⫻ 20 1.032± 0.62% 1.043± 0.67%
FIG. 8. Lateral dose profiles of a 20⫻ 20 cm2 field calculated in the FFF 25⫻ 25 1.037± 0.66% 1.057± 0.66%
machine. The dose, D, was normalized to the central axis dose level, and the 30⫻ 30 1.029± 0.66% 1.061± 0.75%
lateral width x⬘ was normalized to x50. Note that in this case, the depth- 35⫻ 35 1.044± 0.66% 1.072± 0.74%
dependency of the dose inside the field seen in the case with the FF has 40⫻ 40 1.047± 0.70% 1.053± 0.76%
virtually disappeared.

Medical Physics, Vol. 33, No. 6, June 2006

1601 Titt et al.: FFF Monte Carlo simulation
of 4 mm, hence leading to an enlargement of the measured
penumbra. Also, the transport cutoff energy for electrons
contributed to the discrepancy by preventing transport of
electrons below certain energies.
The simulations of the FFF machine showed a maximum
possible dose rate increase on the central axis by more than a
factor of 2 compared to irradiation with the FF. Comparison
of the photon fluences, ␾␥, averaged over the phantom sur-
face, without and with the FF, showed a ratio FFF/FF of 3.52
for a field size of 10⫻ 10 cm2, whereas this ratio was 1.49
for a field size of 40⫻ 40 cm2.
These results are promising in terms of the possibility of
reducing the beam-on time by about a factor of 2 and thus
cutting the exposure of the patient to leakage photons from
the treatment head into half. In small treatment fields deliv-
ered by a conventional accelerator, about 50% of the photons
would be absorbed or scattered in the FF, and add to the
out-of-field dose. With the FF removed, most of these pho-
tons could contribute to the target dose.
The thickness of the bremsstrahlung target 共6-MV mode兲
that is given in the available technical documentation is not
sufficient to completely stop the incident primary electrons.
This can be confirmed by a simple continuous slowing-down
calculation. In a conventional accelerator, these electrons are
blocked by the FF. If the FF is removed, the electron fluence
on the monitor chamber surface increases by a factor of 6.2
based on Monte Carlo simulations 共Table V兲. The increase
poses the risk of placing the chamber outside the range of its
reliable operation. Electron fluence also increases on the
phantom surface 共Table V兲 indicating a potential risk of de-
livering an elevated skin dose as can be seen in Fig. 4. The
number of electrons reaching the phantom strongly depends
on the field size: the calculations show that electron fluence
in the 40⫻ 40 cm2 field is about 6 times higher than that in a
10⫻ 10 cm2 field 共Table V兲. This is because primary elec-
trons leaving the target have a broad angular distribution and,
after further collisions downstream, can be considered as
originating from a volume source located upstream of the
field defining jaws. The energy distribution of electrons
reaching the phantom peaks around 1.6 MeV 共Fig. 5兲. The
shape of the peak is characteristic of the spectrum broaden-
ing of a monoenergetic beam due to ionization collisions
共energy straggling兲. Its location is consistent with the above-
mentioned continuous slowing-down calculation.
On the one hand, these excess electrons should be pre-
vented from reaching the ionization chamber and the phan-
tom; on the other hand, any mechanism used to accomplish
this should avoid attenuation and scattering of the brems-
strahlung photons. Knowing the energy spectrum of elec-
trons 共Fig. 5兲, and therefore their ranges, we chose to inves-
tigate two different electron-stopping foils 共2-mm copper and
11-mm nylon兲.
By introducing the copper foil into the beam, the electron
fluence, averaged over the phantom surface 共50⫻ 50 cm2兲
could be decreased by a factor of 3.1 at the phantom surface,
whereas it was still 28% above the value with the FF in
place. The photon fluence at this location would be 37%
Medical Physics, Vol. 33, No. 6, June 2006
1601
higher compared to the FF mode. With the nylon foil, the
values were somewhat less favorable; the electron fluence
would be decreased by a factor of 2.7, but would still be
about 48% above the flattened beam, whereas the photon
fluence would be 40% higher.
At the downstream surface of the ionization chamber, ␾e
can be reduced by a factor of 5.3, when a 2-mm copper foil
is used; this corresponds to 22% higher than the FF value.
The nylon foil decreased the electron fluence by a factor of
4.4, which corresponds to 46% above the FF value.
These results show that the best choice of electron stop-
ping foils would be the copper foil with a thickness of 2 mm
which would result in normal performance of the ionization
chamber, i.e., there would be no danger of saturation because
the dynamic range of this device can handle a 22% signal
increase.18
The lateral profiles calculated with and without the FF in
the beamline show significant differences 共Figs. 7 and 8兲.
The FFF profiles, in contrast to the flattened beam, exhibit a
strong variation across the field exceeding a factor of 2 in
large fields. The shape of these profiles exhibits a relatively
small variation with depth. The shape of the flattened beam
changes substantially: at shallow depths, it is concave near
the center and becomes convex as the depth increases. The
out-of-field dose calculated without the flattening filter is
smaller outside the field edge for small field sizes. Note that
in the region up to about 30 cm near the treatment field,
scatter from the treatment beam dominates over treatment-
head leakage.19 Hence, a patient would be exposed to signifi-
cantly less radiation to normal tissue.
The calculated total scatter factors, Sc,p, defined as the
dose rate at a reference depth for a given field size 共x
⫻ x cm2兲 divided by the dose rate at the same point and
depth for the reference field size 共10⫻ 10 cm2兲, showed less
systematic variation in the case of an unflattened beam 共FFF兲
compared with the values for the system with FF. Sc,p for the
unflattened beam ranges from 0.927 at 2 ⫻ 2 cm2 to 1.047 at
40⫻ 40 cm2 compared to a range of 0.899 for the 2
⫻ 2 cm2 field size to 1.053 at 40⫻ 40 cm2 for the FF setup.
Fits to the calculated Sc,p data yielded
0.8894 + 0.175 01x
f共x兲 = 共3兲
1 + 0.163 88x
for the FFF beamline shown as the solid curve in Fig. 10 and
0.8552 + 0.1479x
f共x兲 = 共4兲
1 + 0.132 74x
for the FF setup shown as the dashed line in Fig. 10.
The smaller systematic variations in Sc,p for the FFF ma-
chine is mainly due to reduction of scattered photon or head
scatter component of the total scatter factor.
V. CONCLUSIONS
A Monte Carlo model to simulate the 6-MV beam of a
Varian Clinac 2100 with and without a FF has been built,
tested, and validated using experimental commissioning data
measured at M. D. Anderson. Depth-dose profiles, lateral
1602 Titt et al.: FFF Monte Carlo simulation
FIG. 10. Total scatter factor Sc,p as a function of field size. The open squares
show data calculated for a standard Varian Clinac 2100 with FF, and the
solid squares represent the results for the FFF machine. Uncertainties of the
MC simulations show the 1␴ confidence interval. The curves are fits to the
simulation data.
beam profiles, and photon and electron fluences as well as
total scatter factors have been calculated for the FFF ma-
chine and compared to data calculated for a standard treat-
ment head with the FF.
The results of the Monte Carlo simulation suggest that the
removal of the FF could result in higher central-axis dose
rates and, hence, shorter IMRT treatments. The out-of-field
dose 共i.e., head scatter, MLC leakage, etc.兲 to the patient can
be decreased, because the beam-on time of the treatment unit
can be decreased by up to a factor of 2. Close to the treat-
ment field, a significant decrease in out-of-field dose from
in-field-scatter photons could be observed for small treat-
ment fields.
Excess electrons, which would normally be scattered and
stopped in the FF would reach the patient. However, they can
be filtered out with a thin copper foil without significantly
compromising the enhanced photon dose rate.
With the data from the simulations done so far, a
treatment-planning system 共TPS兲 will be commissioned for
an FFF accelerator. The accuracy of the TPS standard dose
calculation algorithms will be investigated and, if necessary,
modifications will be made to ensure reliable results. To a
large extend, the investigation will be based on the Monte
Carlo model of the accelerator. After the TPS is commis-
sioned and fully tested, treatment plans will be developed for
both the conventional and the FFF accelerators. A compara-
tive analysis of the plans will be made in terms of the quality
of deliverable treatment, including an analysis of dose distri-
butions, estimation of risk for late effects and comparisons of
treatment delivery times. Further plans also include a similar
study for the 18-MV beam.
ACKNOWLEDGMENTS
This work was supported in part with a research grant
from Varian Medical Systems. The authors wish to thank the
Medical Physics, Vol. 33, No. 6, June 2006
1602
following members of Varian Medical Systems for their sup-
port during the course of this project: Jyrki Alakuijla, Jeff
Amacker, Yves Archambault, Cal Huntzinger, Scott Johnson,
Armin Langenegger, Stan Mansfield, Richard Morse, and Ri-
chard Stark.
a兲
Author to whom correspondence should be addressed. Electronic mail:
utitt@mdanderson.org
1
J. Deng, S. B. Jiang, A. Kapur, J. Li, T. Pawlicki, and C.-M. Ma, “Photon
beam characterization and modelling for Monte Carlo treatment plan-
ning,” Phys. Med. Biol. 45, 411–427 共2000兲.
2
M. K. Fix, M. Stampanoni, P. Manser, E. J. Born, R. Mini, and P. Rüeg-
segger, “A multiple source model for 6 MV photon beam dose calcula-
tions using Monte Carlo,” Phys. Med. Biol. 46, 1407–1427 共2001兲.
3
K. E. Sixel and E. B. Podgorsak, “Buildup region and depth of dose
maximum of megavoltage x-ray beams,” Med. Phys. 21, 411–416
共1994兲.
4
K. E. Sixel and B. A. Faddegon, “Calculation of x-ray spectra for radio-
surgical beams,” Med. Phys. 22, 1657–1661 共1995兲.
5
C. Guillerminet, E. Duverger, R. Gschwind, L. Makovicka, A. Monnier,
F. Hamon, and C. David, “Theoretical and experimental study of spectral
distortions at the output of an accelerator for medical use,” Radiat. Phys.
Chem. 61, 611–613 共2001兲.
6
S. Zefkili, C. Kappas, and J. C. Rosenwald, “On-axis and off-axis, pri-
mary dose component in high energy photon beams,” Med. Phys. 21,
799–808 共1994兲.
7
P. C. Lee, “Monte Carlo simulation of the differential beam hardening
effect of a flattening filter on a therapeutic x-ray beam,” Med. Phys. 24,
1485–1489 共1997兲.
8
P. F. O’Brien, B. A. Gillies, M. Schwartz, C. Young, and P. Davey, “Ra-
diosurgery with unflattened 6-MV photon beams,” Med. Phys. 18, 519–
521 共1991兲.
9
B. A. Gillies, P. F. O’Brien, R. McVittie, C. McParland, and H. Easton,
“Engineering modifications for dynamic stereotactically assisted radio-
therapy,” Med. Phys. 20, 1491–1495 共1993兲.
10
A. Tsechanski, A. F. Bielajew, S. Faermann, and Y. Krutman, “A thin
target approach for portal imaging in medical accelerators,”Phys. Med.
Biol. 43, 2221–2236 共1998兲.
11
J. L. Robar, S. A. Riccio, and M. A. Martin, “Tumor dose enhancement
using modified megavoltage beams and contrast media,” Phys. Med. Biol.
47, 2433–2449 共2002兲.
12
R. Jeraj, T. R. Mackie, J. Balog, G. Olivera, D. Pearson, J. Kapatoes, K.
Ruchala, and P. Reckwendt, “Radiation characteristics of helical tomo-
therapy,” Med. Phys. 31, 396–404 共2004兲.
13
W. Fu, J. Dai, Y. Hu, D. Han, and Y. Song, “Delivery time comparison for
intensity-modulated radiation therapy with/without flattening filter: a
planning study,” Phys. Med. Biol. 49, 1535–1547 共2004兲.
14
D. W. O. Rogers, C. M. Ma, B. Walters, G. X. Ding, D. Sheikh-Bagheri,
and G. Zhang, “BEAMnrc user manual,” NRCC Report No. PIRS-
0509共A兲 revG.
15
H. G. Hughes, H. W. Egdorf, F. Gallmeier, R. C. Little, R. E. Prael, E.
Snow, L. S. Waters, and M. White, “MCNPX User’s Manual,” TPO-E83-
G-UG-X-00001, Rev. 0 共1999兲.
16
S. Cho, O. Vassiliev, S. Lee, H. Liu, G. Ibbott, and R. Mohan, “Reference
photon dosimetry data and reference phase space data for the 6 MV pho-
ton beam from Varian Clinac 2100 series linear accelerators,” Med. Phys.
32, 137-148 共2005兲.
17
F. M. Khan, The Physics of Radiation Therapy, 3rd ed. 共Lippingcott Wil-
liams and Wilkins, New York, 2003兲, p. 179.
18
S. Johnson, Varian Medical Systems Inc., Palo Alto, California, USA
共private communication兲.
19
M. Stovall, C. R. Blackwell, J. Cundiff, D. H. Novak, J. R. Palta, L. K.
Wagner, E. W. Webster, and R. J. Shalek, “Fetal dose from radiotherapy
with photon beams: Report of AAPM Radiation Therapy Committee Task
Group No. 36,” Med. Phys. 22, 63–82 共1995兲.