See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/12473240

Dose calculations for asymmetric fields defined by independent collimators

using symmetric field data

Article in The British journal of radiology · May 2000

DOI: 10.1259/bjr.73.868.10844866 · Source: PubMed

CITATIONS READS

7 630

4 authors, including:

Ioannis A Tsalafoutas Stylianos Nikoletopoulos

Saint Savvas Hospital IASSO Hospital Athens
90 PUBLICATIONS 1,073 CITATIONS 8 PUBLICATIONS 107 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Stylianos Nikoletopoulos on 21 May 2014.

The user has requested enhancement of the downloaded file.

The British Journal of Radiology, 73 (2000), 403±409 E 2000 The British Institute of Radiology

Dose calculations for asymmetric ®elds de®ned by

independent collimators using symmetric ®eld data
1
I A TSALAFOUTAS, MSc, 2S XENOFOS, PhD, 2A PAPALEXOPOULOS, MSc and
2
S NIKOLETOPOULOS, MSc
1
Medical Physics Unit, Konstantopoulio ± Agia Olga Hospital, 3-5 Agias Olgas, Nea Ionia, 142 33, Athens
and 2Medical Physics Department, Iaso Hospital, 37-39 Ki®ssias, Maroussi, 151 23, Athens, Greece

Abstract. Several methods have been developed for the dosimetry of asymmetric radiation ®elds
formed by independently moving collimator jaws. Three of these methods, based on different
principles and modi®ed to comply with our set of available data, are utilized for the calculation of
asymmetric ®eld dose pro®les. All three methods use output factors and per cent depth doses or
tissue maximum ratios of symmetric ®elds. In the ®rst method, calculation of the off-centre ratio
(OCR) of the asymmetric ®eld is based on the symmetric ®eld from which the asymmetric is
originated, by setting the one jaw in an asymmetrical position. In the second method the OCR of
the symmetric ®eld is used for the OCR calculation of the asymmetric ®eld of the same size;
whereas the third method does not allow for the asymmetric OCR calculation. The results
obtained using data for the 6 MV photon beam of a Philips SL-20 linear accelerator indicate that
both the ®rst and second method can accurately reproduce asymmetric ®eld pro®les from
symmetric ®eld data; the third method does not allow for penumbra reproduction, but it is
accurate at the central part of the asymmetric ®eld. The problems encountered in the application
of the three methods are reported and their accuracy is compared.

Linear accelerators equipped with independent to reproduce asymmetric ®elds produced by a

collimators are capable of producing a wide
variety of off-axis radiation ®elds, formerly
done with the use of shielding blocks. This has
found many applications in optimizing radiation
treatment schemes [1] but has increased the
volume of the basic dosimetric data required
and thereby necessitated modi®cations of treat-
ment planning systems.
The implementation of symmetric ®elds into a
treatment planning system requires quite a large
set of data for various ®eld sizes: per cent depth
doses (PDDs) or tissue maximum ratios (TMRs),
output factors (Sc , Sp), cross-beam dose pro®les
at various depths, etc. Several treatment planning
systems calculate asymmetric ®elds on the
assumption that these are equivalent to blocked
®elds [2]. However, the parameters required for
treatment planning implementation depend on the
algorithm used, and the method of their deriva-
tion is often not explicit [2].
Since asymmetric ®elds are the simplest case of
irregular ®elds, others [1, 3±9] have approached
this problem using appropriate algorithms that
utilize the data available for symmetric ®elds.
Three of these later methods have been chosen
Received 1 March 1999 and in revised form 5 August
1999, accepted 25 August 1999.
Philips SL-20 linear accelerator (Crawley, UK)
located in the Iaso Radiotherapy Department.
The ®rst method, proposed by Kwa and collea-
gues [1], applies to the situation where only one of
the independent jaws is closed down to form an
asymmetric ®eld of smaller width or length than
the original symmetric ®eld. This method uses the
original symmetric ®eld pro®le corrected point-
by-point by a correction factor, and is based on
the principle of the Day's equivalent ®eld
calculation [10]. The second method, based on
work by several authors [3±7], generates asym-
metric ®eld pro®les by computing the off-centre
ratio (OCR) (often referred to as the off-axis
ratio, OAR) of the asymmetric ®eld while using
output factors and PDDs of the equivalent
symmetric ®eld. The third method, proposed by
Khan et al [3], is the simplest, but is limited to the
calculation of dose to a point on the central ray of
asymmetric ®elds. In this study it has also been
applied beyond the central ray of the asymmetric
®eld.
The ®rst two methods require software for their
application, with the ®rst being more demanding,
since interpolations and extrapolations are needed
for point-by-point corrections. We studied the
problems arising from their application and
compared their accuracy.

The British Journal of Radiology, April 2000 403

 I A Tsalafoutas, S Xenofos, A Papalexopoulos and S Nikoletopoulos
Materials and methods
The study was performed on an SL-20 linear
accelerator for the 6 MV X-ray beam. A comput-
erized water tank system was used for dose
measurements (Wellhofer water phantom with WP
700S, v3.20 Nucletron Software; Schwazembuck,
Germany). The measurements were carried out at
source-to-surface distance (SSD) set to 100 cm,
using a 0.125 cm3 ionization chamber.
Data for a full set of measurements for
symmetric ®elds and a number of asymmetric
®elds, including cross-beam pro®les at various
depths, PDDs and output factors, were obtained
during dosimetric calibration of the accelerator.
These data were made available from the WP
700S software, in text ®le format for manipulation
(in 2 mm steps). The asymmetric pro®les were
®elds 666 cm2, 10610 cm2 and 15615 cm2 with
offsets 3 cm, 5 cm and 7.5 cm, respectively, and
666 cm2 with offset 8.7 cm at depths 1.5 cm,
5 cm and 10 cm.
The symmetric pro®les were all taken as relative
cross-beam pro®les with the 100% corresponding to
the centre of the ®eld (x50). To convert these to
absolute dose pro®les, the central axis dose at depth
1.5 cm for the 10610 cm2 symmetric ®eld was used
as reference dose (unit dose or 100%). Thus, the
absolute dose Ds for any symmetric ®eld at any
point and depth is given by the following equation:
Ds WL; d; x†~Sc WL†Sp WL†PDDs WL; d†OCRs WL; d; x†
(1)
where W and L are the width and length of the
®eld (de®ned at SSD5100 cm), respectively, x is
the distance from the central axis, d is the depth,
the subscript ``s'' represents symmetric ®elds and
Sc and Sp are the collimator scatter factor and the
phantom scatter factor, respectively (normalized
to 10610 ®eld).
Method 1
The ®rst method (Kwa et al) has been described
for the case of isocentric technique [1]. To meet
the requirements of the standard SSD technique
used in this study for cross-beam pro®le acquisi-
tion, modi®cations have to be made to the
corresponding equations.
The original equation for the description of the
asymmetric ®eld dose pro®le (Da) can be written
using the same notation as before, as:
Da Wa L; d; x†~Sc;p WL† TMRs WL; d†|
100 OCRs WL; d; x†|
CF Wa L; WL; d; x† 2†
404
where 100 OCRs (WL, d, x) is the relative dose
pro®le and
CF Wa L; WL; d; x†~
s
Sc;p ‰ 2Wa {W {2x†L; dŠTMRs ‰ 2Wa {W {2x†L; dŠ
Sc;p ‰ W {2x†LŠTMRs ‰ W {2x†LŠ
(3)
where Wa and W are the asymmetric and the
corresponding symmetric ®eld width, L is the ®eld
length de®ned at the source-to-axis distance
(SAD), while TMR and Sc,p (5ScSp) refer to
symmetric ®elds. The symmetric ®eld correspond-
ing to a given asymmetric ®eld is that with width
twice that of the longest opening of the asymmet-
rically set jaws and equal length. To illustrate the
formation of an asymmetric ®eld from the
corresponding symmetric ®eld, a schematic repre-
sentation of the beam's eye view of the radiation
®elds is given in Figure 1.
We must note here that Kwa and colleagues
derived Equation (3) considering that the right
collimator was the one that was moving to the
centre forming the asymmetric ®eld. If the left
collimator is forming the asymmetric ®eld (which
was the case in our study) then 22x should be
substituted by +2x in Equation (3).
For the standard SSD method, Equations (2)
and (3) become:
Da Wa L; d; x†~‰ SSDzd max††= SSDzd†Š2 |
S c WL†Sp Wd Ld †|
TMRs Wd Ld ; d†|
100OCRs WL; d; x†|
CF Wa L; WL; d; x†
(4)
where
s
Sc r1 †Sp r1;d †TMR r1;d ; d†
CF Wa L; W L; d; x†~
Sc r2 †Sp r2;d †TMR r2;d ; d†
(5a)
r15 equivalent square of rectangular ®eld
‰ 2Wa {W z2x†LŠ 5b†
and
r25 equivalent square of rectangular ®eld
‰ W z2x†LŠ 5c†
Symmetric and asymmetric ®eld dimensions are
de®ned at the surface and, for the ®eld size at
depth the subscript ``d'' is used (i.e.
Wd5W(SSD+d)/SSD).
The British Journal of Radiology, April 2000
Dose calculations for asymmetric ®elds
W 30630 cm2 are about 1±2% at the shoulder and
L Wa
Figure 1. Schematic representation of the beam's eye
view of an asymmetric ®eld and its corresponding
symmetric ®eld.
Equations (4) and (5) modify the original
Equations (2) and (3), to ®t the case of standard
SSD technique. TMRs were obtained from PDDs
using the following equation [11]:
" #2
PDD r; d; SSD† Sp rd max† † SSDzd
TMR rd ; d†~
100 Sp rd † SSDzd max†
(6)
where d(max) is the depth of maximum dose
(d(max)51.5 cm) and r, rd(max) and rd are the
equivalent square ®eld size at the surface, at depth
d(max) and at depth d, respectively. Symmetric
®eld data (PDDs, output factors, pro®les etc.)
were introduced in a computer program written
for the application of this method, which utilizes
Equations (4) to (6). Interpolation and extrapola-
tion algorithms and penumbral corrections (to be
described later) were also included, so that any
asymmetric ®eld pro®le could be produced from
the corresponding symmetric ®eld data.
The method described above requires the
availability of the OCR of the symmetric ®eld
with the same length as the asymmetric ®eld and a
width twice that of the longest opening of the
asymmetrically set jaws (Equation 4). That is, for
a 15615 cm2 asymmetric ®eld with offset 7.5 cm,
the corresponding symmetric ®eld OCR required
is 30615 cm2. However, data for rectangular
®elds are not always available, whereas data for a
square ®eld with length twice that of the
asymmetric, i.e. 30630 cm2, are more likely to
be available. As shown in Figure 2, the differences
in the relative dose pro®les of a 30615 cm2 and a
The British Journal of Radiology, April 2000
up to 3.5% at the tail. Thus the relative pro®les
(as OCRs) of the corresponding square symmetric
®elds can be used instead. This approximation
reduces the number of pro®les required for the
application of this method, but the accuracy is
slightly reduced.
The problem in the application of this method,
as mentioned by Kwa et al [1], is the penumbra of
the asymmetrically set jaw, which in fact deter-
mines the left ®eld edge. To deal with this they
have suggested that a correction factor should be
set semi-empirically, using as a basis the penum-
bra of the symmetric ®eld. The exact methodol-
ogy, however, was not made explicit. Assuming
that the penumbra of the asymmetric ®eld is
similar to that of the symmetric ®eld with equal
dimensions [1, 8], we used another procedure for
penumbra correction. The pro®le of the sym-
metric ®eld with the same dimensions as the
asymmetric pro®le was offset so as to bring its
centre to the centre of the asymmetric pro®le for
the depth in question. This pro®le was then
renormalized so as to match the dose value 1.5 cm
from the left ®eld edge (towards the central ray)
with the dose value, at the same point, of the
asymmetric pro®les produced. Finally, the left
penumbra of the renormalized symmetric ®eld is
attached to the asymmetric pro®le to form the
new (penumbra-corrected) asymmetric pro®le.
The choice of 1.5 cm is based on the fact that
TMR and Sc,p of ®elds smaller than 363 cm2
(Equation 5a) are not available and cannot be
accurately extrapolated [1].
Method 2
The second method that will be used in this
study for the calculation of asymmetric ®elds was
developed by Loshek and Keller [5], based on
work by Chui and Mohan [4]. In its ®nal
formulation, using the normalization proposed
by Thomas and Thomas [6], this method predicts
the absolute asymmetric ®eld pro®le (Da) by
means of the following equation:
Da WL; d; x†~Sc WL†S p WL†|
PDDs WL; d†OCRa WL;x; d† 7a†
where
POCR x†
OCRa WL; x; d†~OCRs WL; x{x0 ; d†
POCR x{x0 †
(7b)
Sc , Sp and PDDs are the same as in Equation
(1) (i.e. for the corresponding symmetric ®elds),
and OCRa and OCRs are the off-centre ratios of
the asymmetric ®eld (with centre at x5x0) and the
405
 I A Tsalafoutas, S Xenofos, A Papalexopoulos and S Nikoletopoulos
symmetric ®eld (centre at x50), respectively. The
POCR is the primary off-centre ratio, de®ned as
the ratio of dose at a point off the central ray to
the dose at the corresponding point at the central
axis, for an in®nitely large ®eld [6].
Method 3
The third method to be applied is the one
proposed by Khan et al [3] for predicting the dose
along the central ray of an asymmetric ®eld. It is
described by Equation (7a) if OCRa (WL,x,d) is
substituted by POCR(x). As a result, this method
does not allow for penumbra reproduction.
The major dif®culty in the application of the
second and third methods is in the de®nition and
measurement of the POCR [4±6]. In this study, to
simplify the method and avoid additional meas-
urements, POCR was approximated from the
40640 cm2 cross-beam pro®les at depth acquired
during the standard dosimetric calibration of the
accelerator. This approximation is expected to
introduce small errors only: ®rst, the actual ®eld
sizes are usually smaller than 40640 cm2 so that
the effects of the collimators in the POCR are not
important [4]; and second, the errors introduced
by scatter when using the cross-beam pro®les at
depth instead of the corresponding POCR are to
some extent cancelled out in the ratio of Equation
(7b) as they affect both POCR(x) and
POCR(x2x0).
Results and discussion
The pro®les for a 15615 cm2 asymmetric ®eld
at depths 1.5 cm and 10 cm, calculated using the
®rst method, before penumbral corrections, using
406
Figure 2. Relative cross-beam dose
pro®les of the 30630 cm2 and
30615 cm2 symmetric ®elds. d,
depth.
both the 30630 cm2 and 30615 cm2 symmetric
®eld OCRs, are shown in Figure 3. The pro®les
calculated using the OCR of the symmetric
30615 cm2 ®eld present errors only at the left
®eld edge. Those calculated using the OCR of the
symmetric 30630 cm2 ®eld present additional
errors at the right ®eld edge: the errors in the
shoulder are about 1%, while the errors in the tail
are more prominent.
Corrections in the right ®eld edge, when using
the corresponding square symmetric ®eld pro®les,
can be obtained in a similar way to the
corrections of the left ®eld edge, described in
the Materials and methods section. The right
geometrical ®eld edge at the corresponding depth
was chosen as the normalization point, that is the
point where the penumbra of the symmetric ®eld
with equal dimension is attached. This arbitrary
choice was based on the fact that this point is
known a priori for all ®elds, keeping the rationale
of the correction algorithm uniform.
The pro®les for the 666 cm2, 10610 cm2 and
15615 cm2 asymmetric ®elds with offsets 3 cm,
5 cm and 7.5 cm, respectively, as predicted by the
®rst method, using the corresponding square
symmetric ®eld pro®les and the proposed penum-
bral corrections (both left and right edge), as well
as by the second and third method, are shown in
Figures 4, 5 and 6.
It can be seen that both the ®rst method,
corrected as proposed, and the second method
can accurately reproduce asymmetric pro®les.
Some offsets in the penumbra are due to the
fact that experimental points have been taken at
0.2 cm intervals, which may introduce a maxi-
mum positional error of ¡1 mm. It may also be
observed that the third method, even though
The British Journal of Radiology, April 2000
Dose calculations for asymmetric ®elds

Figure 3. Asymmetric cross-beam

dose pro®les of the 15615 cm2
(offset57.5 cm) ®eld, at depths
1.5 cm and 10 cm, calculated with
the Kwa et al method before the
penumbral corrections. The pro®les
have been calculated using both the
30630 cm2 and 30615 cm2 sym-
metric ®eld off-centre ratio (OCR).
d, depth.

Table 1. Dose values expressed relative to the dose at depth51.5 cm of a symmetric 10610 cm2 ®eld, which is
considered equal to 100

Field size (cm2) Depth Data point Kwa et al's Loshek and Khan et al's
(perpendicular ray (cm) method [1] Keller method [5] method [3]
off-axis distance (cm))

666 1.5 99 98.6 98.9 98.9

(3) 5 84.6 84.3 83.9 84.2
10 63.7 63.7 63 63.7
10610 1.5 103.5 103.9 104 104.2
(5) 5 89.3 89.7 89.7 89.4
10 69.5 69.3 69.8 69.6
15615 1.5 108.5 109 109 109
(7.6) 5 94.3 94.6 94.8 94.5
10 74 74 74 73.8

Figure 4. Asymmetric cross-beam

dose pro®les of 666 cm2 (off-
set53 cm) ®eld, calculated by three
different methods. d, depth.

The British Journal of Radiology, April 2000 407

 I A Tsalafoutas, S Xenofos, A Papalexopoulos and S Nikoletopoulos
originally proposed for the central ray of the
asymmetric ®eld, can also be applied with good
accuracy for the central part of the ®eld (about
60% of the ®eld width). The calculated dose
values (by all three methods), for various depths
along a perpendicular ray at the centre of the
asymmetric ®eld, are essentially the same as
shown in Table 1.
It should be mentioned that the ®rst method is
quite demanding, as far as the number of pro®les
required for its application is concerned. While
the data requirements can be greatly reduced by
using square symmetric ®elds and the proposed
right ®eld edge correction procedure, it is obvious
that a different symmetric pro®le is required for
asymmetric ®elds of the same dimensions but with
408
Figure 5. Asymmetric cross-beam
dose pro®les of a 10610 cm2 (off-
set55 cm) ®eld, calculated by three
different methods. d, depth.
different offsets. For example, the asymmetric
666 cm2 ®eld with offset 8.7 cm requires a
23.466 cm2 pro®le, which is unlikely to be
available and has to be interpolated from the
treatment planning system. So the accuracy of the
reproduced asymmetric pro®le depends on the
accuracy of the interpolated symmetric pro®le.
The second method, however, does not present
this problem, as it uses the symmetric ®eld with
the same dimensions as the asymmetric, irrespec-
tive of the offset. In Figure 7, the 666 cm2 ®eld
with offset 8.7 cm has been calculated using the
second and the third method. Again, the asym-
metric pro®les calculated with Method 2 agree
well with those measured, while the third method
is accurate at the central part of the ®eld.
Figure 6. Asymmetric cross-beam
dose pro®les of a 15615 cm2 (off-
set57.5 cm) ®eld, calculated by
three different methods. d, depth.
The British Journal of Radiology, April 2000
 Dose calculations for asymmetric ®elds
From Figures 4±7, it can be seen that the
approximation of POCR with the 40640 cm2
cross-beam pro®les at depth does not evidently
affect the pro®les calculated by the second
method. The differences observed between meas-
ured and calculated pro®les at the edges of the
®eld (Figure 6) could in principle be attributed to
the POCR approximation. If this were true,
however, the errors should increase with depth,
an effect that is not observed. So, although the
POCR approximation introduces a small error,
most of the differences should be attributable to
experimental and positional rounding errors.
Conclusions
The method of Kwa and colleagues, modi®ed
for standard SSD technique and with the addi-
tional penumbral corrections proposed, works
well and exhibits very good accuracy.
However, Loshek and Keller's method, which is
also accurate, has the advantage of using a single
equation for the whole pro®le and directly
producing the penumbra region. It is also less
demanding in the number of pro®le data required.
The approximation of POCR with the
40640 cm2 cross-beam pro®les at depth results
in calculated asymmetric pro®les that compare
well with measured data, and greatly simpli®es the
application of this method.
For point dose calculations at the central part
of the asymmetric ®eld, the prominent simplicity
and good accuracy of the method of Khan et al
makes it the method of choice for straightforward
monitor unit calculations.
The British Journal of Radiology, April 2000
View publication stats
Figure 7. Asymmetric cross-beam
dose pro®les of a 666 cm2 (off-
set58.7 cm) ®eld, calculated by two
methods. d, depth.
References
1. Kwa W, Kornelsen RO, Harrison RW, El-Khatib
E. Dosimetry for asymmetric x-ray ®elds. Med Phys
1994;21:1599±604.
2. International Commission on Radiation Units and
Measurements. Representation and modi®cation of
photon beams. In: ICRU Report 42. Use of
computers in external beam radiotherapy proce-
dures with high-energy photons and electrons.
Bethesda, MD: ICRU, 1987:8±11.
3. Khan FM, Gerbi BJ, Deibel FC. Dosimetry of
asymmetric x-ray collimators. Med Phys 1986;13:
936±41.
4. Chui C, Mohan R. Off-center ratios for three
dimensional dose calculations. Med Phys 1986;13:
409±12.
5. Loshek DD, Keller KA. Beam pro®le generator for
asymmetric ®elds. Med Phys 1988;15:604±10.
6. Thomas SJ, Thomas RL. A beam generation
algorithm for linear accelerators with independent
collimators. Phys Med Biol 1990;35:325±32.
7. Chui C, Mohan R, Fontenla D. Dose computations
for asymmetric ®elds de®ned by independent jaws.
Med Phys 1988;15:93±5.
8. Palta JR, Ayyangar KM, Suntharalingam N.
Dosimetric characteristics of a 6 MV beam from
a linear accelerator with asymmetric jaws. Int
J Radiat Oncol Biol Phys 1988;14:383±7.
9. Gibbons JP, Khan FM. Calculation of dose in
asymmetric photon beams. Med Phys 1995;22:1451±
7.
10. Day MJ. A note on the calculation of dose in X-ray
®elds. Br J Radiol 1950;23:368.
11. Khan FM. A system of dosimetric calculations. In:
The physics of radiation therapy. Baltimore, MD:
Wilkins & Wilkins, 1984:187.
409