Fixed bending current for Elekta SL25 linearaccelerators

J. G .M. Kok*Department of Radiotherapy, University Hospital Utrecht, Heidelberglaan 100,3584 CX Utrecht, The Netherlands.

In a medical linear accelerator a bending magnet is used tobend the electron beam produced by the accelerator tube, in thetreatment direction. For each electron energy the strength of themagnetic �eld has to be set to a speci�c level. Changing themagnetic �eld strength is done by changing the electric currentthrough the bending magnet. When electron energy andmagnetic �eld strength are not matched, performance of thelinac can be affected. As electron energy, magnetic �eld strengthand electrical current through the bending magnet are relatedto each other, it is reasonable to assume that for each electronenergy the correct bending current can be predetermined. Thiscalculated bending current reduces the number of variableparameters used to set up a treatment beam. Predetermining avariable simpli�es the tuning procedures. It also prevents adeviation of the electron beam energy being compensated byvariation of the bending current. Preventing false machinesettings can contribute to increase linac performance andreduce down time and cost of ownership.

Introduction

To obtain the relatively high energies required forradiotherapy treatment, the linear accelerator tubemust have a certain length. Only at the lowest energies(&4 MeV) are standing wave tubes small enough to beintegrated into the radiation head; usually acceleratortubes are mounted in the arm of the treatmentmachine. The orientation of the accelerator tube ismore or less parallel to the patient and the electronbeam produced by the accelerator tube is bent towardsthe patient by means of an (electro)magnet, termed thebending or steering magnet. There are several types ofbending magnets, which are usually speci� ed by theangle of de� ection, for example, the 90 degree, 270degree or slalom magnets [1]. The strength of themagnetic � eld of the bending magnet is set to bendelectrons with a speci� c energy. This means that foreach treatment energy the magnetic � eld strength hasto be changed. This is done by adjusting the electriccurrent through the electromagnet.

The electron beam produced by the accelerator tube isnot mono energetic. Ideally it can be seen as a Gaussiandistribution with a speci�c width. When the electronbeam energy does not match the magnetic �eld

strength of the bending magnet, the electron beam willbe bent in a different direction. It is possible that a partof the electron beam may hit the energy de� ning ormeasuring slits or wall of the bending chamber, causinga part of the electron spectrum to not appear at theelectron window or photon target. This means that onlyelectrons within a certain bandwidth will pass thebending magnet. This bandwidth is called the energyacceptance bandwidth of the bending magnet. Forexample the energy acceptance bandwidth of an ElektaSL25 (slitted) bending system is +9% of the centreenergy (data provided by Elekta, Crawley, UK). Ideallythe energy of the accelerated electron beam is matchedto the magnetic �eld, but in practice this is not alwaysthe case. When there is a large discrepancy of theelectron energy and the magnetic � eld strength, anincrease of linac instability and damage to parts of themachine can be experienced. As long as � eld � atnessand dose rate are within speci� ed limits there is nointerlock to detect poor settings.

To increase linac reliability and reduce down time ofour six Elekta linear accelerators (four SLi15 with sixelectron energies ranging from 4 to 15 MeV and twoSLi20 with the additional 18 and 20 MeV) a procedureof determining the magnetic �eld strength, or thebending current, has been developed.

Although this work is based on experience with theElekta SL25 series linear accelerators, the theory ofmatching electron energy and magnetic � eld strengthis valid for other machine types.

Before an (Elekta) linear accelerator is installed in ahospital it is tested and setup at the factory, so usuallyonly small changes have to be made to comply with thecustomer acceptance test. At the factory the setup startswith the inherited settings of previous machines.Because of spread in components and calibrations,changes have to be made to settings like the bendingcurrent. Usually these changes stay well within thedesign limits. During clinical life the performance ofthe linac is regularly checked and if necessary adjusted.At our department the percent depth dose (PDD) ofthe electron and photon beams is measured everymonth using a water phantom. The stability of the PDDbetween the measurements is better than 0.5 mm [2].When the depth dose curve of a speci� c energybecomes out of tolerance this shift in energy iscorrected by changing the bending current (ormagnetic � eld strength). This is a simple and fast wayto change the energy of the beam. On the other handthe beam bending magnet only acts as an energy

169Journal of Medical Engineering & TechnologyISSN 0309-1902 print/ISSN 1464-522X online Ó 2001 Taylor & Francis Ltd

http://www.tandf.co.uk/journalsDOI: 10.1080/03091900110060758

*Author for correspondence; email: kok@radth.ruu.nl

Journal of Medical Engineering & Technology, Volume 25, Number 4, (July/August 2001), pages 169 – 172

J M

ed E

ng T

echn

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y D

okuz

Eyl

ul U

niv.

on

11/0

5/14

For

pers

onal

use

onl

y.

170

window (� lter), the energy of the electron beamproduced by the accelerator tube remaining the same.Usually energy changes do not occur very frequently.However, after changing components like the magne-tron or electron gun larger changes are expected. Afterseveral years the match between the acceleratedelectron beam and the magnetic � eld strength of thebending magnet can be lost. Two kinds of mismatcheshave been observed.

. The magnetic �eld strength of the bendingmagnet is more or less correct but the energy ofthe accelerated electron beam deviates.

. The energy of the accelerated electron beam isshifted and the magnetic �eld strength is adjustedto compensate for the changed energy.

In both cases the linac can produce a ‘usable’ beam butit is clear that the machine does not perform at itsoptimum. To maintain optimal machine settings overthe lifespan of the linac it is important to stay close tothe appropriate tuning procedures. Usually theseprocedures are quite complicated and time consuming.To simplify tuning it would be useful to have somepredetermined parameters. As the only function of thebending magnet is to bend electrons with a speci�cenergy, it is likely that this parameter (bending current)can be set to a calculated and � xed value.

Methods

The electron beam energy can be de� ned in severalways [3,4]. The depth dose at 50% is a good indicationof the average electron energy E0 but because of theclinical relevance of PDD 80% in our department thisdepth is used for Q.A. of the linac. The de�nition weuse for the PDD 80% is: PDD 80%=‘Nominal Energy’/3,with the energy in MeV and the PDD in cm. Thecalibration of the linacs is checked every month using awater phantom. The measuring system was a Nucletronwater phantom (WMS Nucletron, Veenendaal, TheNetherlands) using diode detectors type P Silicon(Scanditronics Medical AB, Uppsala, Sweden), with a� xed SSD (source surface distance) of one meter. Theaccepted deviation for the PDD 80% is 1 mm. ThePDD 50% used in this study (ideal PDD 50%) is relatedto the PDD 80% by the depth dose curve of eachindividual electron beam. For the same ‘nominal’energy the ratio between the two is almost identical(sd. of the ratio s 0.0031 up to 0.013) and for furthercalculations the average of the six linacs was used. Thecorrection factor given by the NCS report [5] is used tocalculate the average electron energy E0. The process ofdetermining the ideal E0 is illustrated in table 1.

The bending current for each energy was then noted.With a linear regression the relation between themeasured E0 and the measured bending current wascalculated, see table 2.

With these functions the bending current matching theideal E0 was calculated. Most of the calculated and used

bending currents did not diverge more than 1 or 2 percent. Only in four cases the divergence between usedand calculated currents was signi�cant. In � gure 1three examples of measured and calculated data aregiven. From �gure 1a representing machine no. 3 it isclear that there are no large discrepancies between themeasured and the calculated currents. Figure 1b showsthe data of linac 4. Because it was clear beforehand thatthe currents of the 18 and 20 MeV beams were muchhigher than expected, they were not used to calculatethe regression. The apparent reason for these highbending currents are wrong magnetron settings (readradio pulse). The shift in beam energy is compensatedwith the higher bending current. Measurement in thepast [2] have indicated that there are no signs ofsaturation of the magnetic core at these bendingcurrents.

Implementation

The whole idea of �xed bending currents was based onstability problems with the 18 MeV electron beam oflinac number 4. During radiation the linac was

J. G. M. Kok Fixed bending current for Elekta SL25 linear accelerators

Table 1.

Energy PDD 80% Ratio s C1 PDD 50% Ideal E0

4 MeV 13.3 mm 0.825 0.013 16.1 mm 2.46 3.96 MeV6 MeV 20.0 mm 0.834 0.009 24.0 mm 2.39 5.74 MeV8 MeV 26.6 mm 0.847 0.005 31.5 mm 2.35 7.39 MeV

10 MeV 33.3 mm 0.846 0.005 39.4 mm 2.345 9.24 MeV12 MeV 40.0 mm 0.85 0.003 47.1 mm 2.34 11.02 MeV15 MeV 50.0 mm 0.843 0.006 59.3 mm 2.34 13.88 MeV18 MeV 60.0 mm 0.844 0.002 71.0 mm 2.34 16.61 MeV20 MeV 66.7 mm 0.84 0.002 91.4 mm 2.33 18.50 MeV

The ‘nominal’ electron energy, PDD 80% by de�nition, ratiomeasured PDD 50% and PDD 80%, standard deviation of theaverage of six linacs, conversion factor PDD 50% to E0according to NCS report 5 �gure 2, the ideal PDD 50%, theideal E0.

Table 2.

Linac no. Function Variance

1 BC=E066.4+9.6 0.9972 BC=E066.0+10.1 0.9983 BC=E066.5+6.1 0.9994 BC=E067.0+2.3 0.9995 BC=E066.9+9.7 0.9996 BC=E066.4+7.0 0.998

Linac 1,2,5,6, SLi15 with e4 up to e15, linac 3 and 4 with theadditional e18 and 20 MeV beams. The functions forcalculation of the bending current determined by linearregression. (A survey with a current probe (Fluke Y8100)indicates that differences between the functions of the linacsare related to the differences in the calibration of the bendingcurrent of each machine.) The variance indicates the accuracyof the determined linear regression.

J M

ed E

ng T

echn

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y D

okuz

Eyl

ul U

niv.

on

11/0

5/14

For

pers

onal

use

onl

y.

171

frequently interrupted by HT{ and/or DIODE{ over-loads [6]. These problems may have been caused bypoor DC tuning (pulse voltage and current) and ageingof the magnetron, which was 5 years old. When studyingthe magnetron parameters it seemed that it wasrunning at a relatively high power level (higher thanthe power used to produce the 18 MV photon beam). Ifthe power level could be reduced then it was expectedthat the amount of magnetron mis� ring also would bereduced. To � nd a new setting for the 18 MeV electronbeam, � rstly the bending current was altered to thecalculated value. The magnetron frequency and highpower phase shifter (the device that is used to recyclethe RF energy) were then optimized. By changing thebending current, for example 2 amps up and down, itwas checked whether or not the electron beamproduced matched the bending current. If the doserate of the electron beam decreases, with the bending

current set to both approximately 2 A below and abovethe calculated value, the electron energy and themagnetic � eld strength of the bending magnet arematched. To achieve a match for the 18 MeV beam, themagnetron output had to be decreased. The guncurrent was also reduced to match 400 monitor unitsper minute. The magnetron now produced less outputand the overload problems disappeared. After checkingthe PDD, � eld � atness and dosimetry the energy wasready for clinical use. These new settings wereimplemented in September 1998 and the 18 MeV beamperformed well in all respects until 8 December 1999when the magnetron had to be replaced because of ashort circuit in its � lament.

Although the bending current of the 20 MeV beamfrom linac no. 4 also deviates from the calculated value,there were no instability or overload problems. We havetried the new beam setup; although the new settingworked well it has not yet been implemented.

Another energy problem was the 12 MeV from linac no.5. This energy was suffering from dose rate instabilityinduced by a poor frequency response. To � nd a newfrequency-servo setup we started by implementing the

J. G. M. Kok Fixed bending current for Elekta SL25 linear accelerators

{Fault condition of a line type modulator. When the magnetron is poorly matched to themodulator, it can prevent the full recovery of the thyratron. This will induce an overload atthe next charging cycle.{Fault condition of a line type modulator. When the magnetron does not oscillate or arcs itre�ects the HT pulse. This re�ected pulse is discharged via the ‘reverse diode’ to protect themodulator.

Figure 1. (a) Linac no.3, e4, e6, e8, e10, e12, e15, e18, e20, measured and calculated currents are nearly the same. (b) Linacno.4, e4 up to e20, measured bending currents for e18 and e20 are higher than calculated. (c) Linac no.5, e4 up to e15, measuredcurrent for e12 is higher than calculated.

J M

ed E

ng T

echn

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y D

okuz

Eyl

ul U

niv.

on

11/0

5/14

For

pers

onal

use

onl

y.

172

calculated bending current. This energy is now alsoperforming well since the new settings have been used.

Discussion and conclusion

Experience has shown that changing bending currentto the calculated value can speed up tuning procedures.In all cases it must be checked whether the electronbeam energy matches the bending current. This can bedone by varying the bending current and determiningwhether dose rate decreases.

The problems with instability of the 18 MeV electronbeam of linac no. 4 were so serious that replacing themagnetron was considered. With the new settings(calculated bending current) the lifetime of themagnetron is extended. The expenses saved makesthe use of the calculated bending currents attractive.

Determination of the linear regression of the bendingcurrent versus the E0 is simple. Usually the PDD ismeasured frequently and the bending current can bedetermined from the calibration blocks. A scienti� ccalculator or a spreadsheet can be used to determinethe function.

A survey with a current probe (Fluke Y8100) indicatesthat differences between the functions of the linacs arerelated to the differences in the calibration of thebending current of each machine.

Implementation of calculated bending currents forphoton beams can be problematic. A 10 MeV electronbeam does not produce a 10 MV photon beam. To

translate the name energy to the real electron energythe curves of the NACP 43 protocol can be used. Theelectron energy necessary to produce a certain photonbeam can be obtained from the manufacturer of thelinac or by Monte Carlo calculations. It is not knownwhether the complex electron spectrum produced bythe Elekta SL25 type accelerators can introduceproblems.

Acknowledgements

I thank Hans Welleweerd for his suggestions andcomments during the revision of the manuscript.

References

1. GREENE, D., and WILLIAMS, P., 1997, Linear Accelerators forRadiation Therapy. (Bristol, UK and Philadelphia, PA, USA:Institute of Physics Publishing) ISBN 0 7503 0476 6.

2. KOK, J. G. M., and WELLEWEERD, J., 1999, Finding mechan-isms responsible for the spectral distribution of electronbeams produced by a linear accelerator, Med. Phys. 26,2589.

3. ICRU35, 1984, Radiation Dosimetry: electron beams with energiesbetween 1 and 50 MeV (International Commission onRadiation Units and Measurements).

4. KLEVENHAGEN, S. C., 1985, Physics of Electron BeamTherapy. (Adam Hilger Ltd Bristol, UK and Boston, MA,USA) ISBN 0-85274-781-0.

5. NCS report 5, December 1989, Code of practice for thedosimetry of high-energy electron beams. (NetherlandsCommission on Radiation Dosimetry).

6. SIVAS, L., 1994, Microwave Tube Transmitters, (London, UK:Chapman & Hall) ISBN 0-412-57950-2.

J. G. M. Kok Fixed bending current for Elekta SL25 linear accelerators

J M

ed E

ng T

echn

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y D

okuz

Eyl

ul U

niv.

on

11/0

5/14

For

pers

onal

use

onl

y.