Characterization of the output for helical delivery of intensity modulated

slit beams
J. P. Balog, T. R. Mackie,a) P. Reckwerdt, M. Glass, and L. Angelos
Department of Medical Physics, 1530 MSC, 1300 University Avenue, Madison, Wisconsin 53706
~Received 21 November 1997; accepted for publication 27 October 1998!
The UW tomotherapy workbench utilizes a convolution/superposition based dose calculation and
optimization program. It specifies the energy fluence that must be delivered from each leaf for each
phantom projection angle. This requires that the spectrum of the radiation emitted from the one-
dimensional MLC ~multileaf collimator! attached to the linear accelerator be determined. The steps
involved in that process are described. The spectrum along the central axis of the slit beam was
determined, as well as the softening with off-axis position. Moreover, the magnitude of the energy-
fluence output had to be quantified on a per MU ~monitor unit! basis. This was done for a single leaf
along the central axis of the beam. Factors, which modify that energy-fluence output, were inves-
tigated. The output increases with off-axis position due to the horns of the beam. The output for a
leaf of interest will also increase if additional leaves are open due to the absence of the tongue-
and-groove effect and penumbra blurring. The energy-fluence increase per leaf increases by 4.9% if
an adjacent leaf is open. No other factors related to the state of additional leaves were found to
significantly increase the energy-fluence output for an individual leaf. © 1999 American Associa-
tion of Physicists in Medicine. @S0094-2405~99!00301-6#

Key words: MIMiC, convolution/superposition, energy fluence, tomotherapy

I. INTRODUCTION Our unique tomotherapy workbench is first described

A convolution/superposition ~C/S! dose calculation algo-
rithm for an intensity modulating one-dimensional multileaf
collimator ~MLC! has been developed. The MLC used is the
NOMOS MIMiC, or multivane intensity modulating collima-
tor ~NOMOS Corporation, Sewickly, PA!.1 It is a commer-
cially available product, designed as an add on to existing
linear accelerators, used to deliver sequential tomotherapy.2,3
It has been incorporated into a dedicated helical, tomo-
therapy workbench.
Traditionally, the dose calculation from linear accelerator
output to an arbitrary point in a phantom per MU is a func-
tion field size.4 The output increases with the field size due to
an increase in the collimator scatter factor and to an increase
in phantom scatter. Recent studies have shown that the col-
limator scatter factor is mostly due to extra-focal sources of
radiation inside the accelerator head that are allowed to irra-
diate the phantom as the collimating jaws are separated.5 The
chief contributor to this extra-focal radiation is the flattening
filter. The collimator output factor term is a measure of the
increase in photon energy fluence with increasing field size.
The narrow field size for tomotherapy and the leaves them-
selves act as septa to reject photons scattered higher in the
head of the accelerator. The magnitude of the energy-fluence
output when multiple leaves are open is the key to determin-
ing monitor unit calculations for tomotherapy.
Energy-fluence based dose calculation algorithms sim-
plify the dose modeling. As long as the incident energy flu-
ence can be properly modeled, the C/S will inherently ac-
count for the phantom scatter. This is true regardless of the
SSD, the field size, or the depth of measurement.
~Sec. I A!. Next, the differences between the commercial
planing algorithm available for the MIMiC and our newly
developed algorithm are explained ~Secs. I B and I C!. Later,
the methodology involved in deriving the energy-fluence
output from our system is described ~Sec. II!. Finally, the
dosimetric concerns for a helical tomotherapy system are dis-
cussed along with the advantages of an energy-fluence based
C/S planning algorithm for a one-dimensional ~1D! intensity-
modulating slit-beam collimator ~Sec. IV!.
A. The tomotherapy workbench
Tomotherapy employs intensity-modulated slit-beam rota-
tional therapy. The NOMOS MIMiC collimator produces
two rectangular fields of radiation. Intercepting the rectangu-
lar fields are two independent rows of twenty tongue-and-
groove ~T&G! tungsten vanes that are each capable of mov-
ing in and out of the beam path.
Each vane controls a ‘‘beamlet’’ which, ignoring its T&G
projects to 0.97 cm along the long side of the rectangular slit
~fan beam direction, Fig. 1! and 0.82 cm along the narrow
direction ~vertical direction! at 100 cm away ~0.80 cm2 at
100 cm away!. Modulation of the slit beam is achieved by
varying the duration for which a leaf will block the radiation
per arbitrary gantry arc. The radiation beam is modeled as a
number of static intensity-modulated slit beams. Typically
there will be 72 modeled beams, or projections, per rotation,
although different angular resolutions are possible. There-
fore, each required leaf intensity value per projection is de-
livered in a five-degree arc.
The MIMiC has been incorporated into a dedicated tomo-

55 Med. Phys. 26 „1…, January 1999 0094-2405/99/26„1…/55/10/$15.00 © 1999 Am. Assoc. Phys. Med. 55
56 Balog et al.: Characterization of the output for slit beams
FIG. 1. A side view of the tomotherapy workbench geometry, and the direc-
tional terminology used.
therapy workbench that has been assembled to simulate the
characteristics of an actual tomotherapy treatment.6 It con-
sists of an Orion 4 MV linear accelerator ~GE Medical Sys-
tems, Buc, France! permanently directed parallel to the hori-
zon. The modulated radiation strikes a phantom placed on a
computer controlled rotating table ~Compumotor AT6400
3-axis PC-based motion controller, Compumotor Division of
Parker Hannifin Co., Rohnert Park, CA!. The center of rota-
tion on the phantom actuator was set to be 93 cm from the
source. ~This distance allowed the phantom actuator to be
placed between the MIMiC and the focused, post-patient
megavoltage radiation detectors!. The rotating phantom
simulates the rotating accelerator that would be employed by
an actual tomotherapy machine. The phantom can also be
translated vertically to simulate the longitudinal movement
of a patient. Simultaneous rotation and elevation of the phan-
tom enables helical delivery of the slit beam to be achieved.7
We are utilizing only the top row of twenty leaves. This is
all that is needed for helical tomotherapy. Helical tomo-
therapy employs continuous irradiation plus continuous gan-
try rotation and patient translation. Once the phantom has
been initially set up, as many slices as are desired may be
treated without junctioning artifacts.
An illustration of the workbench geometry and directional
terminology is shown in Fig. 1.
B. The NOMOS Peacock planning system
The commercial planning and optimization program ac-
companying the NOMOS system, Peacock, is an inverse sys-
tem in that the desired dose distribution is specified and the
required radiation output is calculated.8–10 Their optimiza-
tion algorithm employs a simulated annealing algorithm.11–13
The forward dose calculation algorithm used in the Pea-
cock system is similar to conventional dose calculation algo-
rithms. It is based on measured dosimetric parameterization
using a finite sized pencil-beam model.14,15 The dose to any
point in a phantom is the sum of the dose that is produced
from each open leaf. The individual leaf calculations incor-
porate an interpolated off-axis ratio appropriate for each leaf
position based on measured data. The values are corrected
for varying depths by interpolated TMR values also derived
from measured TMR values. The calculated dose in a homo-
Medical Physics, Vol. 26, No. 1, January 1999
56
geneous phantom is valid as long as sufficient beam data was
recorded such that the interpolations are accurate. A draw-
back with the Peacock system is that the individual leaf dose
parameterizations do not accurately account for scatter per-
turbation by tissue inhomogenieties. Another drawback to
the Peacock algorithm is that it does not model penumbral
changes during delivery.16,17
C. The convolution/superposition algorithm
We also use an inverse optimization code, but the optimi-
zation is based on the convolution/superposition ~C/S! dose
calculation method.18–20
The C/S predicted nonzero energy-fluence levels are pro-
jected through the phantom and attenuated as they interact
with the phantom. That yields the TERMA distribution
throughout the intersection of the beam with the phantom.
The TERMA distribution is convolved with the dose kernels.
The energy-fluence attenuation and the superposition calcu-
lations are done on a pixel-by-pixel basis. Both calculations
incorporate the radiological path length traversed by the ra-
diation. The pixel densities are inferred from charge-transfer
~CT! data. The convolution code better accounts for the per-
turbation of dose from tissue inhomogeneities.21 A drawback
with the C/S method is that is can be computationally exten-
sive to properly sample phantom/patient space, and it can
require significant amounts of processing time.22
The difference between the desired dose and the calcu-
lated dose is back projected again to modify the original
energy-fluence values. The dose calculation process contin-
ues in at iterative fashion until the desired and calculated
doses agree acceptably well.
Intensity values for each leaf are calculated for all discrete
angular beam projections. The distribution of all required
intensities for all angles is referred to as the energy-fluence
sinogram. It is analogous to the sinogram data collected in
CT. There are 20 leaf indexed intensity values that should be
delivered for all projection angles ~typically 72! for each
complete rotation of the gantry. ~If the required intensity for
any given leaf is zero, than the leaf sinogram value is zero
and the leaf remains closed for that projection.!
The attenuation coefficient and the dose deposition ker-
nels are energy dependent. This necessitates that the spec-
trum of the emitted radiation, as well as the magnitude, are
known for each leaf position. The determination of the mag-
nitude of the energy-fluence output was the main focus of
this research.
D. Energy-fluence dependence on adjacent leaves
The NOMOS MIMiC incorporates a T&G design be-
tween adjacent leaves to minimize the radiation transmission
between closed adjacent leaves. This results in a shared re-
gion between the two leaves that is not irradiated when a leaf
is closed and then opened. This is illustrated in Fig. 2. The
presence of this notch effect, or the absence of it when adja-
cent leaves are simultaneously opened, also affects the
energy-fluence output per leaf. The magnitude of this effect
was studied.
57 Balog et al.: Characterization of the output for slit beams
FIG. 2. An illustration of the tongue-and-groove effect when leaves are
opened sequentially.
It was hypothesized that opening more leaves beyond the
two adjacent leaves next to a leaf of interest ~LOI! might also
increase the energy-fluence output by the LOI. The magni-
tude of this effect as a function of open leaves had to be
determined. It was also theorized that if there were a closed
leaf between a LOI and a group of open leaves, that the
closed leaf would act as a septa and block most of the extra
energy fluence that would otherwise result from the group of
open leaves. This concept is illustrated in Fig. 3.
E. Energy-fluence dependence on adjacent slices
The effect of energy fluence arriving from open leaf com-
binations from off-axis slices also had to be determined. Spe-
cifically, it had to be decided if there was extra-focal radia-
tion from adjacent slices, and if so, how was it affected by
the number of adjacent slices treated. The number of adja-
cent slices treated determines the overall longitudinal length
of the field. Traditionally, the larger the field length, the
greater the extra-focal radiation.
It was hypothesized that there would not be any extra-
focal radiation as a function of longitudinal field length. This
is because the MIMiC always defines the field width along
the longitudinal direction at any instance and can never vary
for a fixed slice width. If there were extra slices treated
above and below the slice of interest, the energy fluence
from them would be the same nature as the energy fluence
defined from the slice of interest. The net effect, should just
be a superposition of off-slice energy-fluence characteristics.
The convolution/superposition-based treatment-planning al-
gorithm would inherently account for this effect of phantom
scatter on dose output.
Medical Physics, Vol. 26, No. 1, January 1999
57
FIG. 3. A top view of six leaves of an MLC are shown. Only leaf ‘‘e’’ is
closed. The scattered photon on the left is able to increase the output under
leaf ‘‘c’’ when the two leaves labeled ‘‘a’’ and ‘‘b’’ are open along with
leaf c. Had leaf ‘‘e’’ been open also, the energy fluence would have been
increased even more. Extra-focal radiation originating from the right-hand
side of the collimator needs leaves ‘‘c,’’ ‘‘d,’’ and ‘‘e’’ to be open simul-
taneously for it to contribute under leaf ‘‘c.’’
II. METHODS
A. Energy-fluence spectrum characterization
The initial energy spectrum was determined by fitting a
measured percent depth dose curve with the predictions of
the convolution method. A PDD curve for a 10 cm by 10 cm
field at 100 cm SSD was measured by an Exradin air equiva-
lent ion chamber ~Exradin Inc. Lisle, IL! in a Solid Water
phantom ~RMI Corporation, Middleton, WI! with no MLC
present. A broad field was initially used because the mea-
surement is much simpler with a larger field. A Monte Carlo
composed spectrum from Mohan23 was first assumed. The
Mohan spectrum was projected, attenuated, and convolved in
a large virtual water phantom to yield a percent depth dose
distribution. The calculated PDD was compared to the mea-
surement and the spectrum was modified until agreement
was achieved.
Next, a measure of the spectrum and magnitude variation
with off-axis position was obtained for the slit beam. A 1 cm
deep (d max for the 4 MV beam! dose profile obtained when
the MIMiC was attached and irradiated through open leaves
was measured with a Scanditronix diode ~model FP1104,
Scanditronix Corp., Sweden! in a water tank. This profile
was used to predict the initial estimate of the ‘‘horns’’ of the
beam. The increase in energy fluence with off-axis position
at this shallow depth was assumed to equal the increase in
58 Balog et al.: Characterization of the output for slit beams
measured dose with off-axis position. An initial off-axis
beam softening was assumed as well from previous models.
The assumed energy fluence was projected, and a dose dis-
tribution was calculated. The magnitude of the horns and the
degree of beam softening were modified in an iterative fash-
ion until the calculated dose matched both the shallow dose
profile and a 30 cm deep ~same measurement conditions as 1
cm! dose profile acceptably well.
Finally, the full-width-half-maximum distance of a pre-
sumed Gaussian focal spot was varied so that the measured
dose penumbra matched the calculated dose penumbra.
B. Energy-fluence output magnitude per single open
leaf
The magnitude of the energy-fluence output had to be
quantified. It was done so on a per leaf basis. The dose at a
depth of 1 cm in a 10 cm by 10 cm field was determined by
a TG21 calibration24 ~performed at a depth of 5 cm, con-
verted to 1 cm by PDD data!. The dose per monitor unit
~MU! for an individually opened central leaf was determined
relative to the 10 cm by 10 cm field by film and diode do-
simetry. The convolution code predicted the dose per energy
fluence to the same point under identical geometrical condi-
tions ~one open leaf along the central axis at d max). The en-
ergy fluence emitted per MU for this leaf is, therefore, equal
to the measured dose per MU divided by the calculated dose
per energy fluence.
Due to the horns, the energy-fluence output is greater for
off-axis positions. The actual energy-fluence per MU ratio
for an arbitrary, single open leaf is the product of the above-
mentioned calculation multiplied by the relative increase in
energy fluence as a function of the leaf’s off-axis position:
~ Energy Fluence/MU! leaf i 5 ~ Energy Fluence/MU!
* Off Axis Factorleaf i . ~1!
C. The tongue-and-groove effect
The absence of the tongue-and-groove effect was mea-
sured by irradiating XV film through two simultaneously
opened adjacent leaves. Twenty MU were applied. The film
was at an SSD of 100 cm and a depth of 1 cm. Since these
two leaves were opened together, there was not a common
region of overlap between them that blocked the radiation.
Forty MU were then delivered to another film when the
same two leaves were opened sequentially, twenty MU for
each open leaf. These leaves would exhibit a T&G effect,
which would manifest itself in the dose distribution.
The resultant profiles were scanned with a laser densito-
meter with a 50 micron spot size, Model SI Personal Dosim-
eter ~Molecular Dynamics, Sunnyvale, CA!. The scanned op-
tical density values were converted to dose by use of optical
density/dose values obtained from irradiated calibration
films. Integration of the dose profiles across the leaf fan-
beam direction ~Fig. 1! yields the profile integral dose ~PID!.
PID5 E D ~ x ! dx. ~2!
Medical Physics, Vol. 26, No. 1, January 1999
58
D. Energy-fluence dependence on additional open
leaves
The extra energy-fluence increase for an individual leaf
due to extra-focal radiation passing through additional open
adjacent leaves was measured with the Scanditronix diode
that was in a solid water mini-phantom. The diode was at a
depth of 1 cm and the phantom was 1 cm wide along the
fan-beam direction. The diode and phantom were precisely
centered in the beam by measuring the collected charge vs
phantom actuator lateral position. ~The phantom actuator
was able to position all phantom/detector experiments to
within 0.1 mm.! The SSD was 93 cm ~the workbench iso-
center!. The increase in detector response as a function of
open adjacent leaves was recorded.
The extra-focal radiation was measured by an additional
method. The Exradin chamber was placed in a solid water
phantom ~20 cm by 20 wide and 5 cm deep! at a depth of 1
cm. The long axis of the chamber was set vertical ~Fig. 1!.
The chamber was placed outside the width of the MIMiC
leaves. The accelerator was turned on. Once the accelerator
reached its peak output rate ~187 MU/min! a central leaf on
the MIMiC was opened. The phantom actuator then moved
the chamber past the open leaf and past the other side of the
MIMiC leaves. ~The direction of travel was the fan-beam
direction.! The accelerator was then turned off. The transla-
tion time was 80 s. The charge collected by the ion chamber
was recorded. The collected charge represents the integral of
the dose profile. The profile integral charge ~PIC! produced
by the single open leaf is defined as
PIC5 E ~ DQ/DX ! * dx, ~3!
where DQ/DX is the charge collected per measurement
point and dx is the ion chamber translation along the fan-
beam direction.
This measurement was relatively insensitive to slight mis-
alignments between the long axis of the MIMiC and the
direction of the ion chamber translation. The long axis of the
ion chamber is 2.1 cm long. The collected charge should not
be a strong function of slight vertical positioning errors since
the ion chamber is wider than the slit beam.
This was repeated for various combinations of open and
closed leaves. An illustration of a PIC measurement is de-
picted in Fig. 4. This measurement strategy produced an in-
dependent measurement of the T&G effect as previously
measured by film. It should also be more accurate whereas
the film dose produced an informative image.
A PIC was obtained with all leaves closed so that the
charge resulting from leakage through closed leaves could be
subtracted. Therefore, the net PIC represents a measure of
the total energy fluence exiting from open leaves of the
MIMiC.
E. Energy-fluence dependence on adjacent slices
The hypothesis that additional slices would not contribute
extra-focal radiation was investigated. The resultant dose
profile from a 5 cm vertical translation of a phantom across
59 Balog et al.: Characterization of the output for slit beams
FIG. 4. A top view of a profile integral charge measurement ~PIC!. An ion
chamber inside a solid water phantom was translated past a stationary fan
beam produced by various combinations of open leaves.
the slit beam was measured by the Scanditronix diode. The
diode was placed in a solid water phantom at a depth of 1.0
cm. The solid water was 20 cm by 20 cm wide and 5 cm
thick. The diode/phantom combination was placed ;9 cm
below the beam central axis. The beam was turned on and
allowed to ramp up to its nominal output rate ~187 MU/min!.
Once the beam output was stable, all the leaves were opened
and the phantom and diode were translated 5 cm vertically
upward in 80 s. The charge collected was recorded. This
collected charge represents the dose that a point in a phan-
tom that started 9 cm below the center of the ;1 cm wide slit
beam and ended 4 cm below the beam center would receive
due to continuous irradiation during that translation. This
measurement was repeated for a 5 cm vertical translation
downwards such that at the end, the diode returned to its
original location. The diode was then offset a few millime-
ters upwards, and another recording was obtained. This pro-
cess was repeated for many offsets ~2 or 4 mm, depending on
the dose profile gradient at the time! until a dose profile of a
5 cm continuous translation of a slit beam was obtained. The
last dose point measured started 4 cm above the slit-beam
profile and ended 9 cm above, such that a symmetrical pro-
file was obtained. The overall length measured was 13 cm.
The central 5 cm passed through the primary beam. We call
such a profile a ‘‘translation profile.’’ The geometrical setup
for the translation profile is shown in Fig. 5.
A profile ~vertical direction! across the static slit beam
was also obtained for all the leaves open. A finite number of
MU were delivered to the diode in a stationary position. The
diode was then moved a few millimeters ~again depending
on the gradient! and another recording was obtained. This
was repeated until a profile of a static slit beam was resolved.
The measured static beam profile was convolved with a 5
cm wide rectangle function sampled every 2 mm, or for 101
data points. The magnitude of the static profile used for this
convolution was scaled by the ratio of the MU delivered for
the 5 cm translation profile ~249!, divided by the MU deliv-
ered for the static profile ~50!, divided by the number of data
Medical Physics, Vol. 26, No. 1, January 1999
59
FIG. 5. An illustration of the phantom/detector location for the first mea-
surement point relative to the vertical direction ~narrow axis! of the slit
beam at the start of the translation profile ~top figure! and at the end ~bottom
figure!.
points sampled in the rectangle function ~101!. The result
was, therefore, the predicted absolute value profile that
would result from a 5 cm, 80 s, translation across the slit
beam without any extra-focal radiation added. ~The convolu-
tion of the static profile with the rectangle function would
not add extra-focal radiation.!
The whole above process was repeated for only two cen-
tral leaves open. Two leaves were used instead of one be-
cause that made the diode alignment with the beam during
vertical translation less critical.
III. RESULTS
A. Energy fluence determination
Figure 6 shows a comparison of the measured PDD to the
predicted PDD ~10 cm by 10 cm field, SSD5100 cm). The
average energy of the 4 MV beam was determined to be 1.36
MV. The agreement between the two curves indicates that
the 4 MV Mohan19 spectrum without modification accu-
rately models the GE Orion 4 MV spectrum.
A few iterations were required to model the off-axis in-
crease in energy fluence and the off-axis decrease in average
energy in order to achieve sufficient agreement between
measured and calculated dose profiles at a shallow depth of 1
cm and a deep depth of 30 cm. The final beam model as-
sumed that the magnitude of the energy fluence increased by
5% at both ends of the 20 cm wide slit beam ~i.e., 0.5% per
cm increase as a function of off-axis distance! as compared
60 Balog et al.: Characterization of the output for slit beams
FIG. 6. The agreement between the measured and predicted PDD curves for
a 10 cm by 10 cm field at an SSD of 100 cm. The depth is measured in the
beam central-axis direction.
to the central axis. The mass attenuation coefficient of the
primary beam at the beam edge increased by 3%, from its
nominal value of 0.061 cm2/g at the beam center ~at the
surface of the phantom!. We also assumed that the increase
in mass attenuation coefficient was linear with the off-axis
distance. The measured and calculated dose profiles along
the fan-beam direction at a depth of 1 and 30 cm are shown
in Fig. 7 for the slit beam with all leaves open. The measured
profiles were obtained by the Scanditronix diode in a water
tank. ~The asymmetry for the measured data is real, but is of
no consequence. The modulation will correct it.!
A Gaussian source shape with a full-width-half-maximum
distance of 1.5 mm yielded the best fit. The penumbra results
were quite insensitive to source size since the collimation
was relatively far from the source and close to the phantom
~collimator end to phantom535 cm). The NOMOS MIMiC
provides very good beam collimation.
Figure 8 shows calculated and measured dose profiles for
every other set of two MIMiC leaves open without altering
FIG. 7. The measured vs the predicted dose profile for all MIMiC leaves
open at a depth of 1 cm and a depth of 30 cm ~scaled to the 1 cm depth
readings!. The measured profiles were obtained with an ion chamber in a
water tank. The profile distances are in the fan-beam direction.
Medical Physics, Vol. 26, No. 1, January 1999
60
FIG. 8. The measured vs the predicted dose profile for sets of two leaves
open at a depth of 1 cm. The measured profile was obtained with an ion
chamber in a water tank. The profile distance is in the fan-beam direction.
the beam parameterization ~measured with the Scanditronix
diode in a water tank!. The agreement of the calculated and
measured profile in Fig. 8 ~and other leaf arrangements not
shown! represents verification that the energy fluence was
properly parameterized.
B. Energy-fluence magnitude
The dose per MU to a depth of 1 cm for a phantom 100
cm away was determined to be 0.866 cGy ~1/21.1%! for a
single leaf aside the central axis. The convolution program
predicted that a fluence of 1.79* 109 photons/cm2 for an area
of 0.80 cm2 irradiated by one leaf, would be required to yield
a dose of 1 cGy at a depth of 1 cm. The fluence per MU per
single open, central leaf is the ratio of the above parameters
~Flu/MU!1 leaf5~0.866 cGy/MU!
*~1.79*109 photons/cGy cm2!*~0.80 cm2!
51.24*109 photons/MU. ~4!
Since the average photon energy is 1.36 MeV at the central
axis, the energy fluence per leaf is 51.69* 109
photons MeV/MU.
Notice that this value is a constant for the accelerator and
MLC and is independent of the phantom material and posi-
tion. The significance is not so much the specific value of
energy fluence per MU indicated above, but rather the
straightforward method of determining the energy fluence
per MU for any accelerator.
Each required energy-fluence-sinogram value determined
by the optimization code is divided by this output to calcu-
late the MU value that must be delivered through each leaf,
or the MU-sinogram data.
C. The tongue-and-groove effect
The resultant dose profiles from simultaneous and sequen-
tial irradiation of two MIMiC leaves are shown in Fig. 9.
There is a pronounced dip in the middle of the sequential
dose profile. This is predominantly due to the tongue region
61 Balog et al.: Characterization of the output for slit beams
FIG. 9. A comparison of dose profiles resulting from simultaneous and se-
quential delivery of 20 MU through two NOMOS MIMiC leaves to film at
a depth of 1 cm in a phantom 100 cm from the source. The profile distance
is in the fan-beam direction.
between the two leaves that is always present to block the
radiation. A component of the decrease in the dose profile is
due to a penumbra effect.16 However, the majority of the
decrease is due to the obvious tongue-and-groove effect.25
The dose difference is as high as 32%, and it has a full-
width-half-maximum distance of 2.1 mm.
The PID values of the dose profiles to film resulting from
simultaneous and sequential leaf openings were obtained.
That represented a measure of the total energy fluence exit-
ing a single leaf the MIMiC. The integral value for simulta-
neous irradiation was measured to be 5% higher than it was
for sequential irradiation.
D. Energy-fluence dependence on additional open
leaves
It is difficult to separate the extra energy fluence striking
an area under a leaf of interest when an adjacent leaf is open
into an extra-focal radiation component and to a lack of the
T&G effect component. Regardless of what the increase is
due to, as long its magnitude can be quantified, it does not
matter. Therefore, the measurements with the mini-phantom
were normalized to the reading obtained when the diode was
centered under three open leaves. This also obviated the con-
cern over whether the mini-phantom should be the width of a
single leaf with or without a T&G.
The increase in energy fluence was measured to be less
than 1% greater for all twenty leaves open compared to when
only three leaves were open. Furthermore, if there was a
closed leaf between the LOI and the other open leaves, the
increase in energy fluence was even less ~,0.3%!. This in-
dicates that there is a slight septa effect demonstrated by
closed leaves.
The translating ion chamber measurements confirmed the
magnitude of the T&G effect measured with film ~Sec. III C,
the film measurements showed the magnitude of the T&G
effect to decrease the energy fluence from each leaf by 5%!.
The net PIC resulting from translation across two concur-
Medical Physics, Vol. 26, No. 1, January 1999
61
rently open middle leaves during an 80 s irradiation was
1.494 nC. The net PIC that resulted when those same leaves
were irradiated successively was 1.424 nC. This, in close
agreement with the film measurements, indicated that the
decreased dose due to the T&G is 4.9% of the total dose
resultant from two leaves without such an effect. The simul-
taneous recording represents two open leaves plus one T&G
effect. ~A full leaf width is defined as the distance from the
edge of one tongue of one leaf to the corresponding tongue
edge of the adjacent leaf.! The sequential reading represents
two open leaves plus two T&G effects. Therefore, the over-
lap blocks 0.070 nC of charge for an 80 s irradiation. The
PIC for one open leaf without a T&G ~if such a situation
were possible! is calculated to be 1.494 nC plus 0.070 nC
divided by two which equals 0.782 nC.
The PIC recorded for all 20 leaves open was 15.78 nC.
This is really from 20 leaves plus one T&G effect ~at the end
of one side of the MLC!. If there were no T&G effect,
the PIC for 20 leaves should, therefore, be
15.78 nC10.070515.85 nC. Odd leaves 1–19, were opened
and a reading of 7.155 nC was obtained. This was added to
the PIC value of 7.159 nC obtained from even leaves 2–20
opened. That sum of 14.31 nC represents 20 leaves plus 20
T&G effects.
The ratio of 20 open leaves with no T&G effects divided
by 20 open leaves with 20 T&G effects515.85 nC/14.31 nc
51.107. There is 10.7% more energy fluence when there are
no T&G effects. ~This assumes there was no extra-focal ra-
diation.! The previous analysis indicated that when two
leaves were opened simultaneously, there was 4.9% more
energy fluence, for the sum of both leaves, than when those
two leaves were sequentially opened. Another way to view
that is there is 9.8% more energy fluence for the sum of one
leaf. That allows the extra energy fluence calculated for all
leaves open to be compared for the case with two leaves
open:
Extra energy fluence per leaf when all leaves open
510.7%,
Extra energy fluence per leaf when two leaves open
59.8%.
It is possible that the difference between these two values
is due to the presence of extra-focal radiation when many
leaves are open. However, even if it is, it is not appreciable
when compared to the value already attributed to the T&G
effect. Therefore, this extra-focal contribution is ignored.
Moreover, as mentioned in the beginning of this section,
there appears to be a septa effect that decreases this extra-
focal contribution if adjacent leaves are closed.
E. Energy-fluence dependence on adjacent slices
The agreement between the measured 5 cm translation
profile and the predicted 5 cm translation profile shown in
Fig. 10, verifies that the moving slit field irradiation is just a
62 Balog et al.: Characterization of the output for slit beams
FIG. 10. The measured 5 cm vertical translation profile vs the predicted 5 cm
vertical translation profile for all leaves open.
convolution of individual slit-beam profiles. Therefore, as
presumed, there is no additional extra-focal radiation from
longer fields.
The static profile was scaled by the MU delivered during
the translation profile irradiation ~249/50! to be at the mag-
nitude if 249 MU were delivered. This scaled static profile is
shown compared to the measured translation profile in Fig.
11 and it illustrates the increased beam-on time needed for
slice-by-slice treatment delivery. Both charge profiles were
produced by a 249 MU exposure. The magnitude of the col-
lected charge for the translation irradiation is much less since
the irradiation was spread over 5 cm. In order for the trans-
lation charge profile to have the same magnitude, there
would have to be much more beam-on time.
Figure 12 is a comparison of the predicted profiles result-
ing from a 5 cm vertical translation and a 3 cm vertical
translation for all leaves open for the same translation speed
as the measured 5 cm translation profile, 16 s/cm. Notice that
the peaks of the two dose profiles are different. The peak
charge predicted for the 5 cm translation is 84.3 nC. The
peak charge predicted for the 3 cm translation is less at 83.1
FIG. 11. The charge collected across all open leaves for 249 MU. In one
case, the ion chamber was translated 5 cm vertically during irradiation, and
in the other it was stationary ~and moved vertically between irradiations!.
Medical Physics, Vol. 26, No. 1, January 1999
62
FIG. 12. A comparison of the predicted profiles resulting from a 5 cm ver-
tical translation and a 3 cm vertical translation for the same translation speed
as the measured 5 cm translation profile, 16 s/cm.
nC. The peak dose for the 5 cm translation has a broader
contribution from the tails of static beam due to scatter in the
phantom. There is also more transmission through the
MIMiC and collimator jaws due to the longer beam-on time.
The dose in the tails of the 3 cm translation profile is also
about 1 nC lower. This 1 nC difference is much more sig-
nificant in the lower dose region. The dose outside the pen-
umbra for the 5 cm translation irradiation is ;3% of the peak
whereas it is only about 2% of the peak for the 3 cm trans-
lation irradiation. Therefore, larger fields limit the ability to
deliver very low dose values just outside the field. This, of
course, is true for conventional treatments as well.
Figure 13~a! is a comparison of the static profiles mea-
sured, all leaves open and two leaves open. The scales were
set to highlight the maximum values of the profiles. Phantom
scatter increases the dose for the case with all leaves open.
Figure 13~b! again shows the profiles along with a leakage
profile, but with the scales set to emphasize the low dose
values outside the primary energy fluence. The dips just out-
side the field are due to the extra shielding along the outside
diameter of the slit in the NOMOS MIMiC.
The comparison of the predicted 5 cm translation profile
for two leaves open to that measured for a 5 cm translation
profile for two leaves open were essentially identical ~as
were the profiles in Fig. 10 for all leaves open!. This agree-
ment offers additional evidence that translation profiles are
just convolutions of instantaneous static profiles. This dem-
onstrates that it is valid regardless of the instantaneous leaf
arrangement.
IV. DISCUSSION
The energy-fluence modeling from a one-dimensional
multileaf collimator is an involved, but an entirely manage-
able problem. The methodology is straightforward and logi-
cal. The energy spectrum is characterized along the entire
slit, and accounts for off-axis softening. Additional dose pro-
files were calculated and measured for various leaf patterns
63 Balog et al.: Characterization of the output for slit beams
FIG. 13. ~a! is a comparison of the static profiles measured in the vertical
direction for all leaves open, and for two leaves open. The scales were set to
highlight the maximum dose values. ~b! A comparison of the three static
profiles measured in the vertical direction, all leaves open, two leaves open,
and no leaves open ~leakage!. The scales were set to emphasize the low dose
values.
after the spectrum and source modeling were completed. The
agreement between these profiles represents verification that
the energy-fluence model is correct.
The output was quantified on an energy fluence per MU
basis for a single leaf aside the central axis. There are only
two factors that significantly change the energy-fluence out-
put, off-axis position, and the presence or absence of the
T&G effect, regardless of the leaf arrangement. The optimi-
zation code iteratively determines the optimum energy-
fluence output from each leaf position independent of those
fluence-modifying concerns other than the increased attenu-
ation of the radiation originating from off-axis positions.
Once an optimum energy-fluence sinogram has been deter-
mined, the specifics of the radiation delivery are weighed.
Each optimized energy-fluence value is converted into the
number of monitor units required to produce it based on the
single quantified energy-fluence/MU leaf parameter. The
MU required for each leaf is then decreased according to the
extra energy fluence that the leaf receives due to its off-axis
Medical Physics, Vol. 26, No. 1, January 1999
63
position. This modified MU-sinogram data is then decreased
again if leaves adjacent to the LOI are open. The energy
fluence, which emanates from each leaf, is inspected. If ad-
jacent leaves are scheduled to be open, there will not be a
T&G effect, and there will be 4.9% more energy fluence for
each open leaf per MU than for a single open leaf ~while
both leaves are open!.
There are no other modifications of the quantified energy-
fluence output. It was determined that any extra-focal radia-
tion resulting from open leaves beyond the two adjacent
leaves is insignificant, at most one percent when all twenty
leaves are open. Furthermore, there does appear to be a slight
septa effect when there is a closed leaf between the LOI and
a group of open leaves. This septa effect decreases the al-
ready small extra energy-fluence contribution from those
open leaves. Therefore, the extra energy fluence beyond the
adjacent leaves is ignored. This greatly simplifies the energy-
fluence modeling. This simplicity extends to three dimen-
sions since there is no extra-focal radiation from adjacent
slices.
It was speculated that the large source-to-collimator dis-
tance, and the narrow slit width were responsible for the lack
of significant amounts of extra-focal radiation. Extra-focal
radiation is, by definition, not focused from the source.
Therefore, only a small contribution has the proper trajectory
to traverse through the narrow aperture.
A clinical tomotherapy machine will not require a flatten-
ing filter. The tungsten leaves will provide the appropriate
modulation of the triangularly peaked beam. Therefore, with-
out this scattering source, there will be even less extra-focal
radiation in the accelerator head, and this should result in
less extra-focal radiation passing through the MLC. There
will also be less beam softening with off-axis position with-
out a flattening filter.
There is, of course, an increase in output to a point under
a leaf of interest as more leaves beyond the adjacent leaves
are opened. This increase is continuous from 1 to 20 open
leaves. It was also demonstrated that there is a dose increase
to a point as more slices are treated. This extra dose is due to
phantom scatter, and not due to extra-focal radiation because
predictions of the convolution/superposition algorithm,
which inherently accounts for phantom scatter, agree with
measurements. Essentially, the C/S algorithm for a 1D MLC
de-couples phantom scatter from extra-focal radiation so that
the two quantities can be determined independently. This
simplification makes the C/S algorithm ideal for a 1D MLC
attached to a linear accelerator.
The net leakage and phantom scatter is integrated for all
slices delivered via moving slit irradiation. This significantly
enhances the overall dose outside the primary energy flu-
ence. Figure 13~b! demonstrates the difference between leak-
age and phantom scatter. The similarity of the two-leaves-
open and the leakage profile outside the primary energy
fluence indicates that there is little phantom scatter away
from the field so the dose is due to leakage. It is this leakage
that is integrated when more slices are treated.
The leakage from off-axis slices should, therefore, be
64 Balog et al.: Characterization of the output for slit beams
made as small as possible. Leakage has been incorporated
into the treatment-planning algorithm.
The difference between the two-leaves-open and all-
leaves-open profile ~Fig. 13! is due almost exclusively to
phantom scatter ~as discussed there is very little extra-focal
radiation!, which cannot be modified except by beam energy.
Presently, the specification for leakage through the pri-
mary collimator and both pairs of jaws for existing linear-
accelerator collimators is 0.1% of the useful beam. The to-
motherapy jaws under development will try to reduce this
amount by a factor of 10 using a thicker primary collimator.
One last point about phantom scatter and leakage integra-
tion should be made. The relative dose outside the primary
energy fluence, compared to the dose inside, is more than a
simple scalar multiple of the number of slices treated times
the static dose outside the primary energy fluence. This is
because the peak dose for a translation profile results prima-
rily from the average dose inside the primary energy fluence.
This is less than the peak dose produced from a static open
leaf. The smaller the slit beam is, the less is the average dose
inside the area irradiated by the slit beam, thereby reducing
efficiency for moving slit-beam delivery.
V. CONCLUSION
The energy-fluence output from a 1D MLC attached to a
linear accelerator can be modeled. The spectrum was deter-
mined along the central axis. The decrease in average
energy-fluence energy with off-axis position was also deter-
mined. The energy-fluence output was quantified per single
open leaf per MU for leaves aside the central axis. Due to the
horns of the beam, the energy-fluence output per MU in-
creases for off-axis locations. The energy-fluence output per
MU for an LOI will also increase if adjacent leaves are open
due to the absence of the T&G effect. It was determined that
there is not significant extra energy fluence if additional
leaves are open. As demonstrated from the insensitivity of
the calculated penumbra on source size, and the lack of
extra-focal radiation, the NOMOS MIMiC provides excellent
beam collimation. There is negligible extra-focal radiation
from off-axis slices. This simplifies the planning process.
The multiplicative nature of the effects of leakage and phan-
tom scatter when many slices are treated was also demon-
strated. Leakage is the limiting factor in the ability of con-
ventional linacs with primary collimators to deliver low dose
values when many slices are treated. Leakage from adjacent
slices should be incorporated into any intensity-modulated
planning system.
ACKNOWLEDGMENTS
This work was funded by a contract with General Electric
Medical Systems and NIH Grant No. CA489.
a!
Electronic mail: mackie@macc.wisc.edu
1 Physicists in Medicine, A Protocol for the determination of absorbed dose
M. Carol, W. H. Grant, III, D. Pavord, P. Eddy, H. S. Targovnik, B.
Butler, S. Woo, J. Figura, V. Onufrey, R. Grossman, and R. Selkar, ‘‘Ini-
tial clinical experience with the Peacock intensity modulation of a 3-D
conformal radiation therapy system,’’ Stereotactic & Functional Neuro-
surgery 66„1–3…, 30–4 ~1996!.
Medical Physics, Vol. 26, No. 1, January 1999
64
2
T. Mackie, T. Holmes, S. Swerdloff, P. Reckwerdt, J. Deasy, J. Yang, B.
Paliwal, and T. Kinsella, ‘‘Tomotherapy: A new concept for the delivery
of conformal radiotherapy,’’ Med. Phys. 20, 1709–1719 ~1993!.
3
T. R. Mackie, T. W. Holmes, P. J. Reckwerdt, and J. Yang, ‘‘Tomo-
therapy: optimized planning and delivery of radiation therapy,’’ Int. J.
Imaging Syst. Technol. 6, 43–55 ~1995!.
4
F. Khan, The Physics of Radiation Therapy ~Williams and Wilkins, Bal-
timore, MD, 1994!, pp. 201–202.
5
T. R. Mackie, H. H. Liu, and E. C. McCullough, ‘‘Calculating output
factors for photon beam radiotherapy using a convolution/superposition
method based on a dual source photon beam model,’’ Med. Phys. 24,
1975–85 ~1997!.
6
J. Balog, Ph.D. thesis, University of Wisconsin at Madison ~1998!.
7
J. Yang and T. R. Mackie, ‘‘An investigation of tomotherapy beam de-
livery,’’ Med. Phys. 24, 425–436 ~1997!.
8
A. Brahme, ‘‘Optimization of stationary and moving beam radiation
therapy techniques,’’ Radiother. Oncol. 12, 129–140 ~1998!.
9
N. H. Barth, ‘‘An inverse problem in radiation therapy,’’ Int. J. Radiat.
Oncol., Biol., Phys. 18, 425–431 ~1990!.
10
M. Goitein, ‘‘The inverse problem,’’ Int. J. Radiat. Oncol., Biol., Phys.
18, 489–491 ~1990!.
11
S. Webb, ‘‘Optimization of conformal radiotherapy dose distributions by
simulated annealing: 2. Inclusion of scatter in the 2D technique,’’ Phys.
Med. Biol. 36„9…, 1227–1237 ~1991!.
12
S. M. Morrill, R. G. Lane, G. Jacobson, and I. I. Rosen, ‘‘Treatment
planning optimization using constrained simulated annealing,’’ Phys.
Med. Biol. 36„10…, 1341–1361 ~1991!.
13
M. P. Carol, ‘‘Peacock™: A system for planning and rotational delivery
of intensity-modulated fields,’’ Int. J. Imaging Syst. Technol. 6, 56–61
~1995!.
14
J. D. Bourland and E. L. Chaney, ‘‘A finite-size pencil beam model for
photon dose calculations in three dimensions,’’ Med. Phys. 19, 1401–
1412 ~1992!.
15
A. R. Bleier, M. P. Carol, B. H. Curran, A. A. Kania, J. D. Scherch, and
E. S. Strenick, ‘‘Dose calculation in Peacock plan,’’ NOMOS Corp. pub-
lication, NOMOS Corp., Sewickly, PA 15143 ~Sept. 1995!.
16
S. Webb and M. Oldham, ‘‘A method to study the characteristics of 3D
dose distributions created by superposition of many intensity-modulated
beams delivered via a slit aperture with multiple absorbing vanes,’’ Phys.
Med. Biol. 41, 2135–2153 ~1996!.
17
M. Oldham and S. Webb, ‘‘Intensity-modulated radiotherapy by means of
static tomotherapy: a planning and verification study,’’ Med. Phys. 24,
827–36 ~1997!.
18
T. Holmes, Ph.D. thesis, University of Wisconsin at Madison ~1993!.
19
T. Holmes and T. R. Mackie, ‘‘A filtered backprojection dose calculation
method for inverse treatment planning,’’ Med. Phys. 21, 303–313 ~1994!.
20
T. W. Holmes, T. R. Mackie, and P. Reckwerdt, ‘‘An iterative filtered
backprojection inverse treatment planning algorithm form tomotherapy,’’
Int. J. Radiat. Oncol., Biol., Phys. 32, 1215–1225 ~1995!.
21
T. R. Mackie, P. Reckwerdt, and N. Papanikolaou, ‘‘3-D photon beam
dose algorithm,’’ in Proceedings from the 1996 AAPM Summer School
~Medical Physics Publishing, Madison, WI, 1996!, pp. 201–222.
22
P. J. Reckwerdt, T. R. Mackie, J. Balog, and T. R. Mcnutt, ‘‘Three di-
mensional inverse treatment optimization for tomotherapy,’’ in Proceed-
ings from the XII International Conference on the Use of Computers in
Radiation Therapy ~Medical Physics Publishing, Madison, WI, 1997!, pp.
420–422.
23
R. Mohan, C. Chui, and L. Lidofsky, ‘‘Energy and angular distributions
of photons from medical linear accelerators,’’ Med. Phys. 12, 592–597
~1985!.
24
Task Group 21, Radiation Therapy Committee, American Association of
from high energy photon and electron beams,’’ Med. Phys. 10, 741–769
~1983!.
25
J. P. Balog, T. R. Mackie, and D. Wenman, ‘‘Multi-leaf collimator inter
leaf transmission,’’ Med. Phys. ~submitted!.