Dosimetric evaluation of compensation in radiotherapy of the breast: MLC intensity modulation and physical compensators

ELSEVIER

Radiotherapy

and Oncology

42 (1997)

249-256

Dosimetric evaluation of compensation in radiotherapy of the breast: MLC intensity modulation and physical compensators Vibeke N. Hansena,*, Philip M. Evansa, Glyn S. Shentall”, Ssrah J. Helyerb, John R. Yarnoldb, William Swindella “Joint Department of Physics, bDepartment of Radiotherapy,

Institute Institute

Received

28 May

of Cancer Research of Cancer Research 1996; revised

version

and the Royal Marsden and the Royal Marsden received

27 November

NHS Trust, Downs Road, Sutton, Surrey, NHS Trust, Downs Road, Sutton, Surrey, 1996; accepted

6 December

SM2 5PT, UK SM2 5PT, UK

1996

Abstract Background and purpose: Electronic portal images may be used to design the compensation required to maximise dose uniformity in the breast from opposed tangential beams. Materials and methods: Four methods of implementing the desired compensation have been studied: a simple wedge, a physical compensator in conjunction with a wedge; one open field plus four shaped multi-leaf-collimated (MLC) fields, and one wedged field in conjunction with three shaped MLC fields. Evaluation was performed using thermoluminescent dosimeters (TLDs) placed inside a phantom which was designed to mimic the human breast. The measured results are compared with both the prediction of the in-house compensation design software and with the dose predicted by the GE Target II planning system. The implications of each method for the time taken to plan and deliver treatment were analysed. Results: The dose inhomogeneity, as measured at seven points in the central plane was greatest for the simple wedge (root mean square (rms) = 4.5%) compared to an open field plus four shaped MLC fields (rms = 2.2%), a wedged field plus three shaped MLC fields (rms = 3.3%), and the physical compensator (rms = 2.4%). The times required to plan and prepare these treatments varied considerably. The standard wedged treatment required under 15 mitt; both MLC-based and the physical compensator treatments required = 50 min. Differences of treatment delivery times were up to 8 min. ConcZusions: These results indicate that the dose inhomogeneity can be reduced by beam intensity modulation designed using EPIDs. 0 1997 Elsevier Science Ireland. Keywords: Breast radiotherapy;

Tissue heterogeneity compensation;

1. Introduction Breast preserving surgery has become the preferred primary treatment for small local breasttumours. It has been

shown that adjuvant radiotherapy after lumpectomy plays an important role both in preventing local tumour recurrence and in improving the disease-freesurvival [7]. For radiotherapy to the breast the most commonly-used technique is the opposedwedged tangential pair. This is adjuvant therapy, as there is no clinical tumour, however, the target volume is well defined as the remaining breast tissue. The ideal dose distribution for the target volume is homogeneous.This is both to achieve uniform biological effect throughout the remaining breast, and to avoid hav* Corresponding

author.

0167-8140/97/$17.00 0 1997 Elsevier PII SO167-8140(96)01895-6

Science Ireland.

All rights reserved

MLC; Intensity modulated beams; Portal imaging

ing volumes of high dosewhich may lead to breastshrinkage, poor cosmesisor other complications. This optimal dose distribution is not achieved by the wedged field technique for most patients [lo]. A nearoptimal dose distribution, that is only limited by the depth dose characteristics of the tangential radiation beams,can be achieved using a physical compensatoror a multi-leaf collimator (MLC). MLCs may be used to achieve intensity modulated beams(IMBs) in two ways: the first is the multiple static field technique of Bortfeld et al. [l], in which dose is delivered in a set of discrete irradiations of different field shapes;the second is the dynamically scanning leaves technique [18], in which leaves are scanned while dose is being delivered. The design of intensity compensation to be delivered using an MLC is discussedin the literature [ 1,4]. Theseintensity

250

V.N. Hansen

et al. /Radiotherapy

modulated beams are largely the result of a more general approach to inverse planning. In a previous paper the use of electronic portal imaging (EPI) to design optimal 2D intensity modulated fields for tangential breast radiotherapy has been described [6]. This paper will concentrate on the practical implementation and dosimetric evaluation of IMBs for the breast, using both the multiple static field technique and physical compensators. The procedures will be outlined and the resultant dose distributions calculated will be compared with measurements. The time implications of these techniques will also be presented.

2. Method 2.1. Optimum beam projile design The physical model used to generate compensators on the basis of electronic portal images (EPIs) has been described in detail [6,13]; hence only a brief summary will be given here. The design of the optimum beam profile consists of three stages: 1. A radiological thickness map of the irradiated tissue within each tangential field is calculated using the intensity information in an EPI. The EPI is corrected for differences in machine output between the field size used to calibrate the detector and that used to acquire the image. It is also corrected for scatter from within the patient [ 151. Once that is done we have a map of the intensity (I> as if the patient were imaged under calibration conditions. This is related to the radiological thickness through Eq. 1

Where Z,, is the intensity measured when no patient is in the field, (Y and fl are calibration parameters and t is the radiological thickness. The thickness, t, can now be found by inverting this equation.

2. This radiological

thickness map is used to derive a pseudo-CT outline set. To do this we make three assumptions: (i) that the breast is symmetrical about the midline; (ii) that the tissue within the radiation field only consists of breast tissue and lung tissue; and (iii) that the outline of the patient where lung is present can be extrapolated from the thickness in regions where only breast tissue is present. Once the outline thickness and the radiological thickness are known the thickness of lung and breast can be estimated [6].

3. The compensated

beam-intensity profile needed to give the most uniform dose distribution achievable (subject to the limitations of the model) is then calcu-

and Oncology

42 (1997)

249-256

lated on a point-by-point basis, we refer to this as the ‘ideal’ beam profile. Firstly for each point in the field the dose delivered, by uncompensated ipsi-lateral and contra-lateral beams is calculated using tissue-maximum-ratio (TMR) tables. The intensity profile required to achieve the most homogeneous dose distribution in the breast tissue is then derived. Once this is done the ‘ideal’ IMB map is turned into a compensator or a set of MLC fields that best approximates to the ‘ideal’ case. We have studied the following pensation: 1. 2. 3. 4.

four methods of com-

Standard wedge Physical compensator in combination with a wedge Four static MLC fields in addition to an open field Three static MLC fields in addition to an open wedged field

For testing our methods we used a pseudo-anatomical phantom. This breast phantom is made from a cast of a large-breasted patient. It consists of an outer layer of beeswax, which is easy to mould, and an inner core of three slabs of ‘solid-water’ each 2 cm thick. In the central slab small inserts for TLD chips were made; in addition there is a lung-equivalent insert in the three slabs of ‘solid-water’. Fig. 1 shows a sagittal and transverse outline of the phantom with seven TLD measuring points indicated in the transverse outline. Each point contains two TLDs. All the points are situated in the central slice, as the phantom could only contain TLDs in the ‘solid-water’ part of the phantom. The plans for the four compensation modalities were generated with the same gantry angles and field size. All dose planning was done using in-house software developed on a VAX 4000160. The results from this will be referred to as ‘VAX calculations’. The dose calculation employed by the VAX calculation is a simple TMR-based dose algorithm, that uses radiological depths, but does not include off-axis factors. This software has also been used to generate dose volume histograms for the various compensation modalities. It is only within the VAX calculation that the whole 3D dose distribution is known, as we have not, yet, a means of putting the pseudo-CT outlines into Target II. We have verified that the compensator design software gives similar results to Target II for the central slice where the outline of the phantom and the insert has been manually traced. When these techniques are to be used on patients we intend to enter the plans into Target II for evaluation, as this is a full planning system with off-axis factors included. We have therefore used Target II to evaluate the results of this present study. Fig. 2 shows the central axis of the ideal intensity map as a solid line; the effect of the breast shape and the lung insert can be seen in the curve. The closer the IMB from

V.N. Hansen

et al. /Radiotherapy

and Oncology

42 (1997)

249-256

251

b)

Bees

wax

Lung "I

Lung

.

*

.

Solid-water

Fig. 1. Diagram showing the anatomical breast phantom in (a) the sagittal plane and (b) transversal plane. The outer contour of the breast-shaped wax volume, the contour of the solid water slabs and the contour of the lung are shown. The positions of the seven measurement (TLD) points are shown in the solid water part of the phantom in the central slice.

any of the four implementations comes to this the better the resultant dose distribution. We now consider each implementation in detail. 2.1.1. Standard wedge Wedges are used for compensation of tangential breast fields in most radiotherapy centres, hence we have taken this treatment as the baseline on which we are trying to improve. The wedge angle was chosen to achieve the flattest possible dose distribution in the mid-plane[6], as this simulates current clinical practice. The phantom was then irradiated according to the planned prescription with the TLDs in place for dose measurement. We have used a Philips SL25 accelerator with a motorised physical wedge filter (i.e., delivering the wedge field as a combination of an open field and a 60” wedged field). The energy used for all treatments was 6 MV. 2.1.2. Physical compensator in combination with a wedge The compensator was constructed using 0.5 mm thick lead sheets, which is radiologically equivalent to 5.7 mm water. The compensator was designed on a 3 x 3 mm grid (The original ‘ideal’ IMB was designed on a 1.5 x 1.5 mm grid equivalent to the pixel size of the imager). The effective attenuation coefficient was measured using the method described by Boyer [2]. The physical compensator was designed to be used in conjunction with a wedge to minimise the number of lead sheets required per field, i.e. the wedge provided a basic level of compensation upon which the physical compensator improved. For the breast

phantom the compensator comprised three layers of individually shaped lead sheets. The software makes a template scaled to the compensator tray. The lead sheets were manually cut out using this template. 2.1.3. Four Static MLCJields in addition to an open jield For MLC intensity modulation calculation the grid size was 1 cm by 1.5 mm, restricted by the physical width of the leaves (1 cm) in one direction and the EPID’s pixel size (1.5 mm.) in the other direction. The third intensity dimension is determined by the number of fields used. We have limited ourselves by a total of five fields from each side, as our planning system GE Target II only allows a total of 10 fields per treatment site. The first irradiation from each side was a standard rectangular field and delivered most of the dose. The four top-up fields were shaped with the MLC and delivered equal, small dose increments. The open field delivered over 80% of the maximum dose. In Fig. 2 the MLC intensity shaped field is shown in dashed line on top of the ‘ideal’ intensity. The MLC shape is made to minimise the difference between the ‘ideal’ dose distribution (known on a 1.5 x 1.5 mm2 grid) and the dose distribution delivered with the MLC. As the width of the MLC leaves are 1 cm, the MLC positions are a mean of the six to seven beam profiles, hence the fit may not be perfect for any one profile. 2.1.4. Three static MLC jelds in addition to an open wedged field The calculation grid was the same as above. The wedge

V.N. Hansen et al. /Radiotherapy and Oncology 42 (1997) 249-256

3. Results 3.1. Dose delivery

60 -5

-4

-3

-2

-1

Position

0

1

2

3

4

5

(cm1

Fig. 2. The ‘ideal’ beam intensity in the central axis of the anatomical breast phantom (solid line), shown together with the MLC leaf setting for the central set of leaves (dotted line). As the MLC leaves are 10 mm wide, their setting is optimised for a set of six to seven ideal intensity beam profiles.

The differential dosevolume histograms(DDVH) of the treatments,ascalculated by the compensationdesignsoftware, are shown in Fig. 3a for the breast volume and Fig. 3b for the lung volume within the treatment field, respectively. We also show the DDVH for the ‘ideal’ IMB and for scanningleaves (seeSection 4). The breast volume of the phantom is taken to be all the solid-water phantom and bees-wax in the volume that is in the irradiated field excluding a l-cm region at the field edge, where the penumbra of the beam is, and excluding 1 cm near the ‘skin’, where the build-up region of the doseis. The narrownessof each peak in Fig. 3a. is a direct measureof the homogeneity of the dose distribution. A delta peak at 100% would mean that all the breast received 100% of the prescribeddose,i.e. the dosedelivered would be absolutely perfect. However, that cannot be achieved even by the ‘ideal’ compensatordue to the dosevariation along a ray-line, i.e. the depth dosevariation from the two oppos-

angle used was the sameasfor the physical compensator. The first irradiation from each side was a standardrectangular field and delivered most of the dosein two segments, one wedged and one unwedged, to produce the required effective wedge angle. This constituted two beamsin the Target II planning system. The three top-up fields were shaped with the MLC and delivered equal, small dose increments. For this method the open/wedged fields deliver over 90% of the dose,i.e. a larger fraction than for the previous method.

W Physical *

Ideal

compensator compensator

2.2. Timing 100

All patients regardless of compensation modality go through a standard treatment planning procedure, where the geometry of the treatment fields is established.At the first treatment sessionan EPI is acquired and from this the compensatoris designed.The actual taking of the EPI does not extend the treatment time provided the EPID is set-up and ready for use. In the time study we only timed the proceduresthat differed between the compensationmodalities. This meansthat we did not time simulation and standardplanning as part of the preparation time and we did not time the patient set-upaspart of the daily treatment time. We timed the design of the compensators,the entering and checking of data on the accelerator control computer, the transfer of the fields to the MLC control computer, manual checks of the MLC fields and the time it took to make the physical compensator.The above mentioned times were all preparation time, only to be done once on each patient. In addition the daily treatment delivery time was measured.These timings were done using a stop watch.

110

120

X Dose

1E 20 GJ n

10

0

90

95

100

105

110

115

X Dose Fig. 3. Differential dose volume histograms (DDVH) for the different treatment strategies discussed in the text, DDVH of breast (a) and DDVH of lung (b).

V.N. Hansen

g”Phys camp.

t open

wedge

t

et al. /Radiotherapy

and Oncology

only

Fig. 4. The normalised dose measured by TLDs at the seven sites (two TLDs at each site) for the four different compensation methods. Here the dose spread can be seen: 1 ((>), 2 (A), 3 (*), 4 ( + ). 5 (cl), 6 (0). 7 ( x ), where the numbers refer to the measurement points shown in Fig. 1.

ing beamslimits the ‘perfect’ compensator.The DDVHs clearly indicate an improvement in the dose homogeneity for all the treatment strategiesrelative to the pure wedged treatment. Also the average dose delivered to the breast volume is reduced, asall plans are normalisedto the ICRU reference point, and for this phantom that is the measurement point 2, which is situated half way between the lung and the skin (2.9 cm from the skin), and at a dept of 6 cm. For the lung volume within the treatment field (Fig. 3b) there is a small reduction in doseto the lung, however, the lung volume will in all five casesreceive between 90 and 110% of the treatment dose, which is well above the tolerance dosefor lung tissue. Hence we do not expect clinical improvements in lung complications compared with a standardwedge treatment.

42 (1997)

249-256

253

The resultsof the TLD phantom dosemeasurementscan be seenin Fig. 4. The thermoluminescenceof the TLDs were measuredusing the NE Technology Rialto reader. The TLDs were previously individually calibrated according to the proceduresdescribedby Mayles et al. [ 111. In Fig. 4 it can clearly be seenthat the simple wedged field gave the largest variation in dose, asmeasuredat the seven sites. The figure showsall the individual TLDs, i.e. two TLD measurementsfor each point, for each treatment modality. Each symbol is a different measurementpoint. In Fig. 5a-d the average of the two TLD readings are shown as a function of the measurementpoints. Error bars of 2% are shown on the measurements,as this was found to be the TLD batch standard deviation. Here the TLD measurementsare shown together with the design software calculation of the dose (VAX calculation) and the dosepredicted by Target II for eachof the sevenpoints. The measurementof point 3 is always 3-4% lower than the dosepredicted by Target II, however, this is what one would expect due to the proximity to the lung (Fig. 1). This is due to lack of scatter from the lung area, which is not modelled in Target II nor in the designprogramme.Fig. 5a showsthe dose for the wedged treatment. Here points 5 and 7, the two points behind the lung, clearly showshigh dose values, as is expected for a purely wedged plan. All graphsin Fig. 5 showsthat the trend of predicted dose is reproduced in the measurements.The measurementsfall within one standarddeviation (2%) for at least five out of the seven points, and of the two points that do not fall within one standard deviation, point 3 is always one of them. In Fig. 6a the % dosemaximum and % dose mini-

01

2345676 Position#

012345678

position #

012345678

Position #

012345678 Position #

-

-Target

..~..Vax

talc.

-*-

Measurements

Fig. 5. The TLD measurements shown as the average of the two TLDs at each measurement predicted by the vax calculation and by Target. (a) Wedge only; (b) open field combined combined with wedge; and (d) wedged field combined with three MLC-shaped fields.

point along with the dose at the measurement points, as with four shaped MLC fields; (c) physical compensator

254

V.N. Hansen

et al. /Radiotherapy

Dose max and min of the 7 points

a)

and Oncology

42 (1997)

249-256

and the measurementsrank the treatment modalities in the same order. The largest difference is seen between the standardwedge treatment and any of the other treatments. These results are summarisedin Table 1. In the last column, data for the whole volume is shown for comparison. In this casethe % rms deviation hasthe sametrend as that calculated on the basis of the seven points in the central slice. Again the conclusions are the same: the doseinhomogeneity is largest for the purely wedgedtreatment. 3.2. Timings

W

Dose spread in the 7 points 14

2 12 28 10 al 6 n% 6 &? 4 2

0 wedge

0 Target

3 MLC + Wedge

H Vax talc.

Phys. Comp.

4MLC + open

E Measurements

Fig. 6. The maximum and minimum dose is shown for each of the treatment modalities for the measurements along with the two sets of calculations. In (a) it can be seen that the minimum does not change significantly, but the change in maximum is significantly reduced by intensity modulation; (b) shows the dose spread for each of the treatment modalities, GE Target II and the vax calculation have similar dose spread predictions.

mum for each of the treatment modalities are shown. The % dose minimum of the seven points do not vary much betweenthe different modalities,but the % dosemaximum is significantly reduced in the three compensatedtreatments. The VAX calculation has a large dose minimum, but a smalldosemaximum, this is probably due to the fact that the VAX calculation assumes the breastphantom to be symmetrical, hence the absolute calibration may be slightly off (for thesecaseswe found a maximum difference in absolutedoseprediction to be 1.O%comparedwith Target II). The total dosespreadas predicted by the VAX calculation andTarget II agreeto within 2%. This is shown in Fig. 6b, where the greater dose spreadin the measurements, due to the uncertainties in the TLD, can be seen. To evaluate the overall result we have calculated the percentageroot mean square(rms) deviation of the dose from the prescribed dose(100%) for each of the different treatment modalities, to seehow they are changedby compensation.The results of this analysisare shownin Fig. 7. Again the standard wedge shows the largest rms. Target

Timings of the procedures that differ from standard treatments are summarisedin Table 2. The compensator design program is developmental software which requires the input of information suchasimage file name,field size and a choice betweentreatment modalities etc. The design takes approximately 3 min. If the modality chosenis either MLC only or MLC combined with wedge, the MLC fields needto be transferred over the network to the MLC contrbl computer (2 min) and then manually checked. If, however, the choice is physical compensator,the compensatorneeds to be manufactured, presently this is done manually and takes approximately 5 min per sheet (for this phantom, three sheetsper field). This processcould be automated using a compensator cutter, saving considerable manpower and time. In addition the field descriptionscurrently have to be typed into the acceleratorcontrol computer, and hencethe more fields the longer the time. Finally, all MLC fields are checked using templatesplaced in the light field of the accelerator. Adding up the preparation time for the different modalities, the simplewedge treatment is clearly the fastest taking 12 min whereasthe other modalities all

% deviation from prescribed dose 6 5

--

4

--

32-

11

+

0 Wedge only

-Target

3 MLC +wedge

-

4 MLC + open

Phys camp.

Vax talc.

-

Measurements

Fig. 7. Comparison of the dose statistics for the four different treatment strategies for measurement and the two calculation methods. Here the %root mean square from the prescribed dose is shown.

V.N. Hansen Table

et al. /Radiotherapy

and Oncology

42 (1997)

249-256

255

1

Summary

of dose homogeneity

Compensators

of the four different

compensators

%Volume 100 f 3%

Wedge alone MLC + open field MLC + wedge field Physical camp Scanning leave ‘Ideal’ compensator

On the basis of the seven measuring

58.2 97.2 93.4 98.8 99.0 100

Measured %rms

Target %llIlS

VAX calculated %lTIlS

4.5 2.2 3.3 2.4 -

4.9 0.98 1.9 1.5

3.5 1.7 1.6 1.7 1.1 0.68

-

The percentage volume within 100 + 3% is derived from the DVH data (calculated), seven TLD points described in the text, or from the whole volume (last column).

take around 50 min for preparation. For the daily treatment times the measured times only involve the actual delivery of the fields, as the patient set-up and gantry positioning will be the same for all modalities discussed here. For each MLC segment it takes an additional 1 min. The main reason for this additional time is internal accelerator checks between each segment. Once dynamic MLC fields become available this additional time is likely be reduced significantly. The use of physical compensators does not increase the treatment time at all. It takes
4. Discussion EPIDs were originally developed for patient positioning measurement [3,5,8]. However, recently EPIDs have been used to extract dosimetric information [9,17] and to design compensators [6,16]. Yin et al. [16] designed physical compensators for AP chest/lung field irradiations. As in this work, they assumed the inhomogeneity was symmetrical relative to the midline. Although they designed a compensator of varying thickness, they only implemented a very simple compensator of two cerrobend blocks of uniform thickness. However, this simple compensator reduced ‘the standard deviation, as measured by TLD from 7.9% to 2.5%. Table 2 Summary

points

Whole volume %rms VAX calculated

5.3 1.3 1.5 0.9 1.6 0.8

whereas the root mean squares (rms) are measured/calculated

for the

The ‘ideal’ breast compensator that has been designed for the breast phantom (Fig. 2) cannot yet be implemented. However, it may be possible to implement this by moulding a high resolution compensator from a metal compound using a compensator cutter. Nevertheless, the results obtained by the VAX calculation of the dose to the whole volume predict that, any of the three other compensation methods has >90% of the breast volume receiving 100% + 3%, which suggests that further improvements may not be needed. In addition none of the methods of compensation had any breast tissue (target) receiving over f5% of the prescribed dose, whereas the pure wedge treatment gave rise to 16% of the breast volume receiving more than f5% of the prescribed dose. Another way of delivering compensation is to use the dynamically-scanning MLC technique [ 181. The DDVH for the breast, resulting from the use of this technique, is shown in Fig. 3a. However, at this stage of development it cannot be put into clinical use. This technique will both overcome the problem of possible instabilities of small segments and the intensity distribution is smoothed rather than stepped. This will enable the production of a dose distribution closer to the ‘ideal’ dose distribution. In most radiotherapy departments the time used on all aspects of planning and delivery of radiotherapy treatment is an increasingly important issue. For this reason we have given times for the different treatment modalities. However, we will not try to judge time against improved dose homogeneity in

of the time study

Design of compensator Enter and check linac data MLC-transfer and checks or fabrication Total preparation time Daily treatment time whole treatment Only procedures time.

that differ

between

of physical

camp.

the implementations

Standard wedge (fin)

Four MLC fields + open field (min)

Three MLC fields + wedged field (fin)

Physical (min)

N/A 12 N/A 12 3

3 30 25 58 11

3 30 17 50 8

3 12 5 mm/sheet 35-55 3

are measured.

The lower

two rows show total preparation

compensator

(ie. 2040

time and daily treatment

min)

delivery

256

V.N. Hansen

et al. /Radiotherapy

this paper. The treatment time is increased by the use of multiple MLC fields. This time study must only be taken as a rough indication of times, as the three new compensation methods are still experimental. Also technical improvements, such as the use of dynamically scanning leaves or automation of the checking procedures of the MLC fields will reduce times considerably. A clinical trial to evaluate these breast compensation methods is being planned. The intention is to limit the trial to large breasted patients, where the dose inhomogeneity with current technique has been shown to be largest [ 12,141 and the radiotherapy complications are most severe. The phantom used in this study was a mould of a large breast, and the dose measurements did show a significant improvement in the dose homogeneity for the IMB, which should lead to a lower complication rate.

5. Conclusions We have designed breast compensators on the basis of EPIs. The practical implementation has been investigated and three different treatment strategies have been tested on a purpose built breast phantom. All compensation strategies showed dosimetric improvements compared to the standard wedged treatment. In terms of time, however, there is a penalty for the dosimetric improvements. This extra time could, however, be reduced significantly with technical improvements and routine use. Acknowledgements

We are grateful to Dr. S. Webb for comments on the manuscript and to Nigel Harper, Stephannie Reise and Colin Nalder for help with the TLD calibration. This work is supported by the Cancer Research Campaign (CRC). References [l]

Bortfeld, T.R., Kahler, D.L., Waldron, T.J. and Boyer, A.L. X-ray field compensation with multileaf collimators. Int. J. Radoat. Oncol. Biol. Phys. 28: 723-730, 1994. [2] Boyer, A.L. Compensating filters for high energy X-rays. Med. Phys. 9: 429-433, 1982. [3] Boyer, A.L., Antonuk, L., Fenster, A., van Herk, M., Meertens, H., Munro, P., Reinstein, L.E. and Wong, J. A review of electronic portal imaging devices (EPIDs). Med. Phys. 19: 1-16, 1992.

and Oncology

42 (1997)

249-256

[4] Convery, D.J. and Rosenbloom, M.E. The generation of intensitymodulated fields for conformal radiotherapy by dynamic collimation Phys. Med. Biol. 37; 1359-1374, 1992. [5] Evans, P.M., Gildersleve, J.Q., Morton, E.J., Swindell, W., Coles, R., Ferraro, M., Rawlings, C., Xiao, Z.R. and Dyer, J. Image comparison techniques for use with megavoltage imaging systems. Br. J. Radiol. 65: 701-709, 1992. [6] Evans, P.M., Hansen, V.N., Mayles, W.P.M., Swindell, W., Torr, M. and Yarnold, J.R. Design of compensators for breast radiotherapy using electronic portal imaging. Radiother. Oncol, 37: 43-54, 1995. [7] Fisher, B., Redmond, C., Poisson, R., Margolese, R., Wolmark, L., Fisher, E., Deutsch, M., Caplan, R., Pilch, Y., Glass, A., Shibata, H., Lemer, H., Terz, J. and Sidorovich, L. Eight-year results of a randomized clinical trail comparing total mastectomy and lumpectomy with or without irradiation in the treatment of breast cancer. New Engl. J. Med. 320; 822-828, 1989. [8] Gildersleve, J., Dearnaley, D.P., Evans, P.M., Law, M., Rawlings, C., and Swindell, W. A randomised trial of patient repositioning during radiotherapy using a megavoltage imaging system. Radiother. Oncol. 31: 161-168, 1994. [9] Hansen, V.N., Evans, P.M. and Swindell, W. The application of transit dosimetry to precision radiotherapy. Med. Phys. 23 (5) 1996. [lo] Mayles, W.P.M., Yamold, J.R. and Webb, S. Improved dose homogeneity in the breast using tissue compensators. Radiother. Oncol. 22: 248-251, 1991. [ll] Mayles, W.P.M., Heisig S. and Mayles, H.M.O. Treatment verification and in vivo dosimetry. In: Radiotherapy Physics in Practice, pp. 227-252. Editors: J.R. Williams and D.I. Thwaites. Oxford University Press, Oxford, 1993. [12] Moody, A.M., Mayles, W.P.M., Bliss, J.M., A’Hem, R.P., Owen, J.R., Regan, J., Broad, B. and Yarnold, J.R. The influence of breast size on late radiation effects and association with radiotherapy dose inhomogeneity. Radiother. Oncol. 33: 106-l 12, 1994. [13] Morton, E.J., Swindell, W., Lewis, D.G. and Evans, P.M. A linear array scintillation crystal-photodiode detector for megavoltage imaging. Med. Phys. 18: 681-691, 1991. [14] Neal, A.J., Torr, M., Helyer, S. and Yarnold, J.R. Correlation of breast heterogeneity with breast size using 3D CT planning and dose-volume histograms. Radiother. Oncol. 34: 210-218, 1995. [15] Swindell, W. and Evans, P.M. Scattered radiation in portal images: A Monte Carlo simulation and a simple physical model. Med. Phys. 23: 63-73, 1996. [16] Yin, F.F., Schell, M.C. and Rubin, P. A technique of automating compensator design for lung inhomogeneity correction using an electronic portal imaging device. Med. Phys. 21: 1729-1732, 1994. 1171 Yin, F.F., Schell, MC. and Rubin, P. Input/output characteristics of a matrix ion-chamber electronic portal imaging device. Med. Phys. 21: 1447-1454, 1994. [18] Yu, C.X., Symons, M.J., Du, M.N., Martinez, A.A. and Wong, J.W. A method for implementing dynamic photon beam intensity modulation using independent jaws and a multileaf collimator. Phys. Med. Biol. 40: 769-787, 1995.