Phys. Med. Biol.

1989, Vol. 34, No 3, 369-378. Printed in the U K

Scient@ note

Is the fictitious virtual source a redundant concept therapeutic electron beams?

G A Sandisonit and W Hudat

for scanned

t Department of Medical Physics, Manitoba Cancer Treatment and Research Foundation, Winnipeg, Manitoba, R3E OV9, Canada $ Department of Physics, University of Manitoba, Winnipeg, Manitoba, Canada
Received 19 January 1988, in final form 1 November 1988

1. Introduction

The fictitious virtual source is an empirical aid to treatment planning but does not have any firm definition. Most commonly it is defined such that, over a limited range, the inverse square law may be applied to predict beam output with actual source-surface distance (Briot and Dutreix1976, Cecatti et a1 1983). Alternatively, but less frequently, the fictitious virtual source location may be determined from the full width at half maximum ( F W H M ) of a beam profile (Schroder-Babo 1983). The location of a fictitious virtual source is problematical since it is not fixed but depends on many parameters (Sweeney et a1 1981), most notable of which are field size and electron beam energy. The empirical method normally used to locate the fictitious virtual source position is a linear extrapolation of data to the source-chamber distance (SCD) axis in a plot of the square root of the inverse of an ionisation reading on the central axis of the beam against SCD. Normally the linear extrapolation is performed over a range of distance large compared with the range of SCD for which the data are collected. The fictitious virtual source position determined by this method moves toward the reference measurement point (typically 100 cm SCD) as the electron beam energy is decreased for a fixed field size and similar behaviour is exhibited for a fixed beam energy as the field size is decreased. Fermi-Eyges transport theory (Rossi and Griesen 1941, Eyges 1948) has previously beenappliedtotherapeutic electronbeamstopredictbeam output with changing patienttreatmentgeometry(Mills et a1 1982, Meyer et a1 1984, Mills et a1 1985). However,these approachesdonotseparatethe influence of backscatterfromthe collimation system intothemonitorchamber(PattersonandShragge1981),which cannotbepredicted by Fermi-Eygestheory,fromtheinfluence of lateralscatter disequilibrium in the electron field due to thecollimation system, which can. Magnetically scannedacceleratorelectronbeams have recentlybeenmodelled in airusing Fermi-Eygestheory by a collimatedisotropicsource(Huizenga and Storchi 1987, Sandison and Huda 1988). The purpose of the present paper is to use the collimated isotropic source model to evaluate the predictive power of Fermi-Eyges theory for both beam output and beam profile F W H M with changing beam energy and S S D . This evaluation uses rectangular fields defined by the photon collimation system alone. Additional collimation is usually introduced for patient treatment and therefore the procedure outlined in this paper is not appropriate for direct application to treatment planning. 0031-9155/89/030369+ lOS02.50 0 Publishing 1989 IOP Ltd 369

370
2. Theory

G A Sandison and W Huda

The spatial distribution of electron fluence F(z, x, y ) for a rectangular field produced by magnetic scanning may be modelled by a collimated isotropic source and is computed on the basis of Fermi-Eyges theory (Sandison and Huda 1988) as

(1) where z is the distance in air from the isotropic point source located at the accelerator exit window, a ( z ) is the half-field width at distance zdefined by the collimation system in the x direction, b(z) is the half-field width at distance z defined by the collimation system in the y direction, x and y are lateral coordinates, a, is the root mean square spatial spread of a pencil beam at distance z when the pencil beam is defined at the level of the x collimation, and a>, is the root mean square spatial spread of a pencil beam at distance z when the pencil beam is defined at the level of the y collimation. The appropriate value of c, (and similarily ay)may be computed from the following relation (Huizenga and Storchi 1987, Sandison and Huda 1988) k(z - 2,)' 6 where f is an empirical fitted constant, k is the linear angular scattering power of air at the beam energy used and z, is the distance from the source to the level of the x collimation. The effect of additional scattering by the monitor chamber at typical treatment distances is negligible. This can be understood by imagining the monitor chamber to be represented as a thin metallic foil surrounding the isotropic source. The foil-covered source still emitselectronsisotropically and at large distances downstream appears identical in effect to an uncovered isotropic source. 2.1. Variation of output with distance in air from the isotropic point source Fermi-Eyges theory alone cannot account for the variation of beam output with field size since it does not consider the component of output variation due to backscatter (BS) from the top of the collimation system to the monitor chamber (Patterson and Shragge 1981), or the small component of output variation due to forward scatter from the collimation system (Lax and Brahme 1980). However, this isotropic source model employing Fermi-Eyges scattering theory is able to account for lateral scatter disequilibrium ( L S D ) effects at therapeutic beam energies. If the collimator scatter is assumed to be a small effect compared with the backscatter component then the output factor (OF) in air a distance z from the source, on the central axis of a rectangular field, is given by

where the subscript r represents the parameters defined in equation (1) for the reference field and

Scanned therapeutic electron

beams

37 1

In addition, the reference field size may be chosen to be a very large square field (36 x 36 cm) forcing the error function values to unity and the referencesourcechamber distance may be set equal to 100 cm. Equation (3) then becomes

fixed field size setting ( a x b ) cm2, the output Equation (3) indicates that for a factor O F varies with distance in air z and lateral scatter disequilibrium LSD, since the ratio BS ( a , b)/ss (a,, b,) is constant. Any geometric dependence of OF due to forward scattering off the photon collimation system has been ignored. Lateral scatter disequilibrium for a rectangular field occurs when thefield penumbra width is greater than or approximately equal to half the corresponding field dimension (e.g. a 6 U, and/or b 6 U , , ) . This results in the LSD term in equation (5) decreasing below unity and reduces the fluence on the centralaxis of the beam below that expected on the basis of the simple inverse square law.

2.2. Variation of

FWHM

with distance in air from the isotropic point source

A further consequence of lateral scatter disequilibrium is the increase in full width at half maximum ( F W H M ) of the x axis profile because of the decreasing central axis fluence value F ( z , 0,O) for the off-axis ratio F ( z , x, 0)/ F(z, 0,O)

The value of the F W H M is given by 2x, when x is the solution to equation (6) for the ratio F(z, x, O)/F(z, 0,O) equal to 1/2. This value of x is obtained iteratively from the equation

3. Materials andmethod The radiation output factor is defined here as the ratio of the ionisation in air on the central axis of a field-of-interest at 100 cm source-chamber distance (SCD) to that of a reference field size. The reference field size adopted in this work was a 36 x 36 cm2 square field. For fixed field size settings, the variation of ionisation readings on the central axis of the beam with SCD are normalised to the 100 cm SCD reading. All in-air ionisation measurements were performed using a Capintec Farmer type PR-06C ionisation chamber of 0.65 cm3 sensitive volume, inner axial length 2.2 cm and inner diameter of 0.7 cm. Selected beam energies of 7, 10, 16, 22 and 32 MeV were investigated for square field sizes in the range 3 X 3 cm to 36 X 36 cm2 and SCD in the range 80-120 cm for fields formed by the photon collimation system (figure 1). In the smallest fields investigated there is a rapid variation in intensity of the beam across the active volume of the chamber. This leads to an underestimation of the ionisation ratio in the worst case of 7.5% which occurs in a 5 X 5 cm2 field at 80 cm S C D for a 7 MeV beam (figure 3(a)). The stem effect measured for the chamber was less than 0.4%. These types of errors are not accounted for in the present data.

372

G A Sandison and W Huda

Quadrupole magnets
l

[XI=
B

wlndow
I

Prlrnary j cofllrnator'

Photon colflmators

105 cm

1
Isocentre

I
Figure 1. Schematic of the Sagittaire Therac 40 linear accelerator showing the magnetic scanning quadrupole and photon collimation system.

In-air beam profile measurements were obtained perpendicularly to the central axis of the therapy electron beam using Kodak Industrex M film pinned to a wooden box frame. The lowest dose rate available for the accelerator was employed (-50 cGy min") to ensure a uniform fluence distribution on the film for the scanned beam. For convenience, the film was retained within the manufacturer's opaque paper cassettewhenused. An opticaldensity-dosecalibrationcurvewasobtainedatthe depth of dose maximum in a tissue equivalent phantom composed of Temex (Stacey et a f 1961), for each electron beam energy. The film was found to be linear within2% for doses in the range 5-30 cGy. To ensure greatest sensitivity in this linear region for penumbra measurements, the maximum of each profile was made to correspond to an optical density in the range 1.5-2.0. Films from the same batch were used in each specified experiment and developed under identical conditions using hand processing. An unexposed film was included with each batch to assess the background fog level. The films werethenreadmanuallybyaSargent-Welchdensitometer(Densichron model PPD) using a 1 mm diameter light aperture.
4. Results
4. l . Fixed jield size setting

The effects of lateral scatter disequilibrium ( L S D ) on beam output in air were investigated by fixing the field size setting ( a x b ) cm2. Equation (3) shows that measurement of the variation in ionisation along the central axis of a small square field with SCD, normalised to the value obtained at100 cm SCD, may be used to investigate deviations from the inverse square law due to Figure 2 shows the results of such measurements LSD. along the central axis of a 6 X 6 cm2 field (defined at 100 cm S C D ) with S C D for beam

Scanned therapeutic electron

beams

373

1.6

a .c

1.2

0.8
0.4

80

85

90

95

100 110
Source

105 distance ( c m 1

115

120

- chamber

Figure 2. Variation of ionisation readings with SCD on the central axis of a 6 cm x 6 cm field at beam energies of 7 ( 0 ) , 16(0), 22(A) MeV. Brokencurveindicatestheinversesquarelawvariation. Full curvesare theoretical predictions including lateral scatter disequilibrium.

energies of 7, 16 and 22 MeV. The predicted inverse square law varation is shown as a broken curve and the full curves indicate the predicted variation using equation (3) with lateral scatter disequilibriumeffects included. The experimental data for32 a MeV beam is not shown in figure 2 but, as is expected, follows the inverse square law over the SCD range of measurements. The 22 MeV beam is predicted to depart from an inverse square law variation only at extended SCD (> 100 cm) due the value of LSD falling below unity. The 16 MeV and 7 MeV beam predicted variations deviate from the inverse square law overthewhole SCD range of measurements.There is good agreement between the experimental and predicted variation especially at extended SCD. The measured variation in ionisation ratio along the central axis of a 7 MeV beam with SCD for square field size settings of 5 x 5 cm2, l o x 10 cm2 and 15 x 15 cm2 are compared in figure 3(a) to the theoretical predictions of equation ( 1 ) which are shown as full curves. The predicted inverse square law is shown as a broken curve. Similar measurements were repeated for a 10 MeV beam and the results are shown in figure 3 ( b ) . Notice that all predicted curves in figures 3(a, b ) deviate from the inverse square law except for the 10 MeV beam with 1 5 X 15 cm2 field size at values of SCD below about 110 cm.
4.2. Variable jield size setting

The beam output factor OF was measured in air at 100 cm SCD for square field sizes in the range 4 x 4 cm2 to 36 x 36 cm2 at electron beam energies of 7, 10, 22 and 32 MeV. These results are shown in figure 4 and demonstrate that the reduction in output with decreasing square field size is strongly dependent on beam energy. Thecollimated isotropic source model employing Fermi-Eyges scattering theory can only predict the inverse square law and lateral scatter disequilibrium effects. The full curves in figure 5 show the predictions of Fermi-Eyges theory in air at 100 cm SCD for the variation in the output factor OF with square field size for the beam energies of 7 and 32 MeV. Differences between the measured and predicted data of figure 5 are due to electrons

374
2.8 1

G A Sandison and W Huda

la I

100 80

95 85

90

105 115

110

120

Source-chamber dlstance ( c m 1
2.4

2.0

1.6
0 .c

e
0

5 .c
.C
I

1.2

0.8

0.4

I
80

,
85
90

95 105

100

110

115

120

Source-chamber distance ( c m l Figure 3. Variation of ionisation readings with S C D for ( a ) 7 MeV beam ( b ) 10 MeV beam on the central axis of square fields of 5 X S cm2 ( O ) , l o x 10 cm2 (0) and 15 x 15 cm2 ( A ) . Broken curve indicates the inverse square law variation. Full curves are the theoretical predictions including lateral scatter disequilibrium.

backscattering from the photon collimators and entering the monitor chamber plus any differences in scatter off the photon collimators reaching the ionisation chamber.
4.3. Beam projile
FWHM

A further effect of lateral scatter disequilibrium is to increase the full width at half maximum(FWHM) of anormalisedbeam profile abovethatpredicted on purely geometrical grounds. Table 1 presents measurements of beam profile F W H M at 100 cm SSD for square fields formed by the photon collimators. The field size is set according to the light field measurements and is varied over the range 3 x 3 cm to 20 x 20 cm2 for beam energies of 7 , 16, 22 and 32 MeV. Notice that at the smallest energy and field size the measured F W H M of the normalised beam profile is over twice as large as would be expected based on the geometrical set-up. Table 1 also compares the measured data with the FWHM of the beams predicted using equation (7).

Scanned therapeutic electron beams

375

loo

0 1

168

12

20

24

28

32

36

Square field length lcml

Figure 4. Variation in output at 100 cm source-chamber distance with field size for beam energies of 7 MeV (0), 10 MeV ( A ) , 22 MeV ( 0 ) and 32 MeV (W). Field is defined by the photon collimators. Broken curves act only as a guide to the eye.

loo
/

o !

O
I 1

12

16

20

3 24 2 length Icm)

28

36

Squarefield

Figure 5. Variation in output at 100 cm SCD with field size for beam energies of 7 MeV (0)and 32 MeV (H). Theoretically predicted curves shown as full curves. Broken curves act only as a guide to the eye.

5. Discussion

The prediction of beam profile shape is an important component of treatment planning. Knowledge of the change in beam outputwith treatment geometry is equally important. Both are related tolateral scatter disequilibrium. Figure 2 shows that for afixed square field size of 6 x 6 cm2 and beamenergies of 16 and 22 MeV the change in beam output with SSD is predicted within 270, by the inverse square law and lateral scatterdisequilibrium effects predicted by equation (3) at values of SCD between 100 cm and 120 cm. The correspondence between the measured and predicted data is equallygoodfor other field sizes greater than 6 X 6 cm2, as demonstrated in figure 3 ( a ) for a 7 MeV beam and figure 3 ( b ) for a 10 MeV beam. The agreement between the measured and

376

G A Sandison and
FWHM

W Huda

Table 1. Comparison of measured values of

Field size (cm?) at 100 cm
SSD

(cm) with predicted values for square field beam profiles.

Electron energy (MeV) 7

16 22 32

M
20.1 15.3 10.9 8.4

P
20.1 15.3 11.1 8.5 7.6 7.0 6.2

M
20.0 15.0 10.0 6.2 5.3 4.5 3.8

P
20.0 15.0 10.0 6.1 5.3 4.5 3.8

M
20.0 15.0 10.0 6.0 5.0 4.1 3.3

P
20.0 15.0 10.0 6.0 5.0 4.1 3.3

M
20.0 15.0 10.0 6.0 5.0 4.0 3.1

P
20.0 15.0 10.0 6.0 5.0 4.0 3.0

20 x 20 1 5 x 15 l o x 10 6x6 5x5 4x4 3x3

7.7
7.1 6.6

M, measured (error k0.2 cm); P, predicted.

predicted data although good is not perfect, especially for the 7 and 10 MeV beams at field sizes less than or equal to 6 X 6 cm2. These discrepancies between prediction and measurement can in part be explained by the variation in intensity of the beam across the chamber in small fields but may also possibly derive from assumptions in the accelerator model used becoming less valid at small values of SCD, beam energy and field size. Namely that the effect of the monitor chamber is negligible, or the implicitassumptioninequation (3) of ignoringthesmallvariation with z in the contribution of forward scatter from the collimation system to the ionisation chamber. Figure 4 shows that the reduction of beam output with a reduction in field size is strongly dependent on beam energy. Figure 5 compares the measured output variation for the 7 MeV and 32 MeV beams with that predicted by equation (3). The discrepancy between the measured and predicted data in figure 5 indicates there is another effect, in additionto lateral scatter disequilibrium, which reduces beam output with a reduction in field size. This is the effect of radiation backscattering from the topof the collimator system into the monitor chamber and any collimator scattereffects into the ionisation chamber. The gun current and hence the electron output is controlled by a feedback mechanism based on the signal from the monitor chamber. As the field size is reduced there is an increase in collimator surface available for the electrons to backscatter into the monitor chamber. This causes an increase in the signal and therefore a corresponding decrease in the actual electron output (Patterson and Shragge 1981). Often noted clinically is the increase in F W H M of normalised electron beam profiles above that normally expected on geometrical grounds. This increase is most noticeable for low beam energies and small field sizes. Table 1 demonstrates that if the values of a , ( z ) are predicted theoretically or measured, then equation (7) may be used to accurately predict the FWHM of rectangular beam profiles for any energy and field size. A comparison of the measured and predicted results clearly shows that the increase in FWHM of beam profiles abovethatgeometricallyexpectedcanbesatisfactorily explained by lateral scatter disequilibrium effects. Agreement between the predicted and measured results for the beam output at a fixed field size and the F W H M data is achieved simply by correctly parametrising the beam in air on the basis of Fermi-Eyges theory (Huizenga and Storchi 1987, Sandison and Huda 1988) and then accounting for lateral scatter disequilibrium effects. These results obviate the concept of a fictitious virtual source for magnetically scanned broad electron beams since both variations in beam output due to lateral electron disequili-

therapeutic Scanned

electron beams

377

brium and beam profile shape theory.

(FWHM)

can be predicted on the basis of Fermi-Eyges

6. Conclusion

The application of Fermi-Eyges theory to electron beamsin air is capable of predicting the changesinbeam output for a fixed field size with SCD or beam energy due to lateral scatter disequilibrium effects. However, overall changes in beam output with field size arenotaccuratelypredicted dueto Fermi-Eygestheorybeingunable to account for the backscattering of electrons from the top of the photon collimators into themonitorchamberorforward scatteringfrom the collimatorstotheionisation chamber.Theincreasein FWHM of beam profiles above the F W H M expected on geometrical grounds atsmall field sizes and low beam energiesis satisfactorily explained by lateral electron disequilibrium effects and is adequately predicted using the collimatedisotropicsourcemodel of scanningaccelerators.Theaccuracy of thebeam output and FWHM predictions demonstrates that the use of a fictitious virtual source concept for magnetically scanned therapeutic electron beams is unnecessary.

Acknowledgments We would like to take this opportunity to express our thanks to the referees for their very helpful comments on this manuscript. This work was supported by a grant from theManitobaHealth ResearchCouncil and in part by agrantfromtheNational Cancer Instistute of Canada.

References
Brahme A 1983 Geometric parameters of clinical electron beams Acra Radial. Suppl. 364 11-9 Briot E and Dutreix A 1976 Dosimetre des Faisceaux delectrons de haute energie dun accelerateur lineaire J. Radial. Electrol. 57 447-54 Cecatti E, Goncalves J , Cecatti S and de Penha Silva M 1983 Effect of the accelerator design on the position of the effective electron source Med. Phys. 10 683-6 Eyges L 1948 Mutliple scattering with energy loss Phys. Ret.. 74 1534-5 Huizenga H and Storchi P 1987 The in-air scattering of clinical electron beams as produced by accelerators with scanning beams and diaphragm collimators Phys. Med. Biol. 32 1011-29 ICRU 1984 International Commission on Radiation Units and Measurements, Radiation Dosimetry, Electron beams with Energies between 1 and 50 MeV ICRU Report 35 (Bethesda, MD: ICRU) Lax I and Brahme A 1980 Collimation of high energy electron beams Acta Radial. Oncol. 19 199-207 Meyer J, Palta J and Hogstrom K 1984 Demonstration of relatively new electron dosimetry measurement techniques on the mevatron 80 Med. Phys. 11 670-7 Mills M, Hogstrom K and Almond P 1982 Prediction of electron beam output factors Med. Phys. 9 60-8 Mills M, Hogstrom K and Fields R 1985 Determination of electron beam output factors for a 20-MeV linear accelerator 1985 Med. Phys. 12 473-6 Patterson M and Shragge P 1981 Characteristics of an 18 MV photon beam from a Therac 20 medical linear accelerator Med. Phys. 8 312-8 Polman H and van der Linden P 1987a Determination of three parameters describing the uncollimated electron beam in air Phys. Med. Biol. 32 345-53 -1987b Description of the uncollimated electron beam in air by means of a directional pencil beam model Phys. Med. Biol. 32 355-63 Rossi B and Griesen K 1941 Cosmic ray theory Ret.. Med. Phys. 13 4250-70

378

G A Sandison and W Huda

Sandison G and Huda W 1988 Application of Fermi scattering theory to a magnetically scanned electron linear accelerator Med. Phys. 15 498-510 Schroder-Babo P 1983 Determination of the virtual electron sourceof a betatron Acta Radio/. Suppl. 364 7-10 Stacey A, Bevan A and Dickens C 1961 A new phantom material employing depolymerised natural rubber Er. J. Radio/. 34 510-5 Sweeney L, Gur D and Bukovitz A 1981 Scatter component and its effect on virtual source and electron beam quality Inr. J. Radial. Onco/. Bioi. Phys. 7 967-71