Monte Carlo simulation of the photoneutron field in linac radiotherapy treatments with different collimation systems

INSTITUTE OF PHYSICS PUBLISHING Phys. Med. Biol. 49 (2004) 571 582 PHYSICS IN MEDICINE AND BIOLOGY PII: S0031-9155(04)69196-2 Monte Carlo simulation of the photoneutron field in linac radiotherapy treatments with different collimation systems A Zanini 1, E Durisi 2,FFasolo 2, C Ongaro 2,LVisca 3,UNastasi 4, KWBurn 5, G Scielzo 6, J O Adler 7, J R M Annand 8 and G Rosner 8 1 INFN Turin, Via P Giuria 1, 10125 Turin, Italy 2 Experimental Physics Department, University of Turin, Via P Giuria 1, 10125 Turin, Italy 3 ASP, Villa Gualino Viale Settimio Severo 65, 10133 Turin, Italy 4 S. Giovanni A S Hospital, Via Cavour 31, 10123 Turin, Italy 5 ENEA, Via Martiri di Monte Sole 4, 40129 Bologna, Italy 6 IRCC Institute for Cancer Research and Treatment, Strada Provinciale 142, Km. 3.95, 10060 Candiolo Turin, Italy 7 Institute of Physics, University of Lund, Lund S223 62, Sweden 8 Department of Physics and Astronomy, University of Glasgow, Glasgow G12 8QQ, UK E-mail: zanini@to.infn.it and durisi@to.infn.it Received 8 September 2003 Published 30 January 2004 Online at stacks.iop.org/pmb/49/571 (DOI: 10.1088/0031-9155/49/4/008) Abstract Bremsstrahlung photon beams produced by linac accelerators are currently the most commonly used method of radiotherapy for tumour treatments. When the photon energy exceeds 10 MeV the patient receives an undesired dose due to photoneutron production in the accelerator head. In the last few decades, new sophisticated techniques such as multileaf collimators have been used for a better definition of the target volume. In this case it is crucial to evaluate the photoneutron dose produced after giant dipole resonance (GDR) excitation of the high Z materials (mainly tungsten and lead) constituting the collimator leaves in view of the optimization of the radiotherapy treatment. A Monte Carlo approach has been used to calculate the photoneutron dose arising from the GDR reaction during radiotherapy with energetic photon beams. The simulation has been performed using the code MCNP4B-GN which is based on MCNP4B, but includes a new routine GAMMAN to model photoneutron production. Results for the facility at IRCC (Istituto per la Ricerca e la Cura del Cancro) Candiolo (Turin), which is based on 18 MV x-rays from a Varian Clinac 2300 C/D, are presented for a variety of different collimator configurations. (Some figures in this article are in colour only in the electronic version) 0031-9155/04/040571+12$30.00 2004 IOP Publishing Ltd Printed in the UK 571

572 A Zanini et al 1. Introduction Nowadays, radiotherapy treatments with static fields defined by traditional collimators or the multileaf collimator (MLC) technique are still widely employed with a linac producing bremsstrahlung photon fields with energy greater than 10 MeV. Although intensity-modulated radiation therapy (IMRT) provides a useful tool in the treatment of various pathologies, this technique uses, in general, a low energy photon beam whose energy is not greater than the (γ, n) threshold energy (e.g., head and neck tumours with 6 MV photon beam; see Verellen and Vahhavere 1999). Neutrons are produced by photonuclear reactions when the energy of the incident photon is higher than the threshold energy of the (γ, n) reaction. This threshold depends on the atomic number of the target: for high atomic numbers it is around 8 MeV whilst for low atomic numbers the threshold is higher (16 MeV for oxygen, 18 MeV for carbon). Therefore linacs with photon energies in the range of 18 25 MeV can produce undesired fast neutrons, both in the accelerator head and directly in the patient s body, which give a non-negligible contribution to the total dose. However, since peak (γ, n) cross sections for high Z materials are around 50 times higher than for low Z ones (W: 400 mb; C: 8 mb), the accelerator head provides the major contribution. The photoneutron energy spectrum is characterized by an evaporation peak in the range 200 700 kev and a relatively weak (10% of the integrated intensity) direct-reaction component in the several MeV energy range. Thus most of the neutrons have maximum biological effectiveness, as pointed out in the publications no 60 and 74 of the International Commission on Radiological Protection (ICRP) in which new radiation weighting factors w R are recommended (ICRP 1991, ICRP 1995, BCRU 1997). Furthermore, since multileaf collimation techniques give a more precise definition of the treatment volume, the gamma dose to the tumour may be raised to improve the effectiveness of the treatment. However, increasing the number of monitor units (MU, see section 3.2) in a programme of treatment will also increase the secondary neutron dose and if therapy is to be optimized this must be quantified. Experimental evaluation of the neutron field at the patient plane is difficult due to the high fluence of photons with respect to neutrons and the pulsed nature of the beam, which causes pile-up in active radiation detectors and noise problems. Therefore the neutron field has been calculated using the Monte Carlo technique. To this end a new photoneutron event generator GAMMAN (Ongaro et al 1999) has been inserted into MCNP4B (MCNP4B 1997), which is widely used for photon and neutron transport calculations. The resultant MCNP4B-GN is capable of detailed modelling of the geometry and materials of a therapy facility, even down to the leaves of an MLC. Calculations are presented for 18 MV x-rays from a Varian Clinac 2300 C/D at IRCC Candiolo in Turin for different MLC configurations. 2. The Monte Carlo simulation and the code MCNP4B-GN A new routine GAMMAN has been developed for the evaluation of (γ, n) processes in high Z elements for photon energies up to approximately 25 MeV, i.e. modelling the giant dipole resonance (GDR) reaction. GAMMAN has been implemented in the Monte Carlo code MCNP4B and the resulting code MCNP4B-GN (Burn and Ongaro 2002) can thus treat the electromagnetic cascade and the photoneutron production and transport. The patch that converts MCNP4B to MCNP4B-GN is available on request from the authors. Recently, in a collaboration with the Karolinska Institute, Stockholm, (γ, n) reactions on low Z elements (Gudowska et al 1999) have been inserted. Benchmarking is currently under way and results will be reported in a future paper.

MC simulation of the photoneutron field in linac radiotherapy treatments 573 The relatively simple photoreaction model assumes that the dominant neutron emission mechanism is evaporation after equilibrium is reached in the excited nucleus. This results in a Maxwellian neutron energy distribution and an isotropic directional distribution. A small (10%) direct neutron knockout component is also considered where the neutron energy is given by E γ S n (S n is the separation energy; E γ is the incident photon energy) and the directional distributionby(a + b sin 2 θ), where a and b are semiempirical parameters and θ is the angle between the direction of the incident photon and the emitted photoneutron. Both (γ, n) and (γ, 2n) channels are treated and since photonuclear cross sections are a factor of 100 lower than for atomic processes (photoelectric, Compton, pair production), they are preferentially weighted in the MC algorithm in order to reduce the statistical uncertainty associated with the neutron responses. Thus MCNP4B-GN simulates photoneutron production in accelerator components (Ongaro et al 2000) and retains the ability of MCNP4B to transport neutrons or photons through complex geometries, which is vital for a realistic modelling of a medical facility. It has been tested against a dedicated measurement of W(γ, n) production yields (Akkurt et al 2003) made at Max Lab in Lund, Sweden, where monochromatic tagged photons (E γ = 13 33 MeV) were incident on 4 8 mm thick sheets of W and the neutrons were detected in a time-of-flight spectrometer. The results from the measurement and calculation were in general agreement within the 10 35% experimental uncertainties. For the calculations described in this paper, the photons responsible for neutron production originate at bremsstrahlung events after the electron beam impinges on the target. The electrons, with their appropriate energy, direction and spatial distributions, are the source particles in the Monte Carlo simulation. The bremsstrahlung model employed in MCNP4B is derived from Koch and Motz theory, revised by Berger and Seltzer, whilst electron transport is treated with a multiscattering model based on Goudsmit Saunderson theory. For these and other details of the physics and data employed see MCNP4B (1997). 3. Results The object of the present study is the Varian 2300 C/D 18 MV x-ray beam installed at IRCC Candiolo in Turin which is equipped with a moveable-jaw photon collimator as well as an MLC. The accelerator geometry has been accurately simulated as shown in figure 1. When the electron beam strikes the tungsten target, a bremsstrahlung photon beam is produced. The primary tungsten collimator then selects the maximum photon angle. A conical shaped iron flattening filter is placed in front of the collimator to reduce the on-axis intensity (i.e. to give a flatter profile) and harden the beam. Below the flattening filter the tungsten secondary collimators ( jaws) define the field in the X and Y directions. Further collimation is provided by a set of 40 tungsten leaf pairs to produce an irregular (tumour) shaped field. The leaves are stepped to create an overlap and thus limit radiation leakage. Photoneutron energy distributions and ambient dose equivalents (source surface distance (SSD) = 100 cm) have been calculated at the patient plane for three distances (3 cm, 8 cm, 15 cm) from the isocentre of the photon field (figure 2). Three different photon field configurations have been simulated: 1. A 10 10 cm 2 square field defined by: jaws 10 10 cm 2,MLC40 40 cm 2. 2. A 10 10 cm 2 square field defined by: jaws 40 40 cm 2,MLC10 10 cm 2 (figure 3). 3. A clinical field configuration (figure 4), typical of what would be used in actual treatment. The jaws define a 10 10 cm 2 square field as in 1 and the 40 leaf pairs of the MLC define the irregular shape corresponding to the cross section of the desired treatment volume.

574 A Zanini et al Figure 1. Geometrical model of the Varian 2300 C/D 18 MV x-ray beam simulated with MCNP4B-GN. Figure 2. Layout of the positions at which neutron dose was calculated for a 10 10 cm 2 photon field and a source surface distance (SSD) of 100 cm. 3.1. Neutron energy spectra at the patient plane Since neutron spectra are affected strongly by bulk moderating materials the concrete walls, floor and ceiling of the treatment room and the treatment table have been included in the computer model in order to take into account the backscattering effects. For photoneutron simulation an energy cut-off of 7.12 MeV is applied to both photons and electrons, below which neutron emission from any element is energetically impossible. The neutron emission threshold is highly dependent on the nuclide considered, being 7.42 MeV for tungsten and 18.7 MeV for carbon. The MCNP results are per starting particle, electrons in this case. Hence to relate the simulation data to the quantities used in a radiotherapy treatment, a normalization factor has been calculated by assessing the photon energy deposited in a water phantom, inserted in the geometrical model, at d max, which is at 3 cm depth for an 18 MV beam.

MC simulation of the photoneutron field in linac radiotherapy treatments 575 Figure 3. MLC cross section. Field edge defined at the patient plane: 10 10 cm 2. Figure 4. MLC cross section. Field edge defined by MLC: clinical configuration. In figure 5 the photoneutron energy spectra are shown, in terms of neutron fluence per 1 Gy photon dose (calculated as described above), for three distances from the isocentre (figure 2). The uncertainty is stochastic and represents one standard deviation. For this case collimator configuration 1 was used. The neutron spectra are characterized by an evaporation component, peaking in the range 200 700 kev, and a direct component above 2 MeV. As would be expected, the latter is most evident at 3 cm distance from the isocentre which is inside the collimated field where energies

576 A Zanini et al Figure 5. Neutron fluence at the patient plane calculated with MCNP4B-GN at different positions with respect to the axis; x-ray field 10 10 cm 2. Figure 6. Neutron fluence at 3 cm from the isocentre (in field) with different collimation systems, calculated with MCNP4B-GN. will be higher. These spectra confirm that most photoneutrons are produced at energies where the biological effectiveness is high. Figures 6, 7, 8 compare neutron spectra for collimator configurations 1, 2 and 3 (see above), at distances 3, 8 and 15 cm from the isocentre, while figure 9 displays the dependence of the integral neutron fluence on collimator configuration and distance from the isocentre. The spectra shown in figures 6 8 result from the photoneutron production and the neutron interactions with the accelerator head. In order to explain the variations in neutron spectra between the different configurations, the neutron production in the components of the accelerator head has been calculated and shown in table 1. The target, primary collimator and flattening filter geometry do not change and thus their contribution remains constant, while the neutron production by photon collisions is higher in the jaws than in the MLC in all the configurations.

MC simulation of the photoneutron field in linac radiotherapy treatments 577 Figure 7. Neutron fluence at 8 cm from the isocentre with different collimation systems, calculated with MCNP4B-GN. Figure 8. Neutron fluence at 15 cm from the isocentre with different collimation systems, calculated with MCNP4B-GN. In configuration 1 the photon beam, collimated by jaws, does not hit the MLC directly; therefore only the scattered photons, degraded in energy and intensity, that strike the MLC contribute to the photoneutron production. In configuration 2 both the collimators ( jaws and MLC) roughly equally contribute to the neutron production. In fact in this case a large direct photon beam hits the MLC set to define a small field (10 10 cm 2 ). In configuration 3 the direct photon beam impinges on the jaws and then on the MLC shaping the photon field in a clinical configuration. In each case the neutron fluences and spectra at the patient plane are the result of photoneutron production (a greater production of course increases the neutron fluence), and nuclear interactions such as scattering and absorption in the accelerator components that

578 A Zanini et al Figure 9. Comparison between neutron integral fluences at different positions with respect to the axis, for three different collimation systems, calculated with MCNP4B-GN. Table 1. Neutron production in photon collisions with various components of the accelerator head, calculated with MCNP4B-GN. Configuration 1 Configuration 2 Configuration 3 Target 15.2% 15.2% 15.2% Flattening filter 8.9% 8.9% 8.9% Primary collimator 44.3% 44.3% 44.3% Jaws 31.0% 16.7% 30.7% MLC 0.3% 14.6% 0.6% Minor components 0.3% 0.3% 0.3% tend to reduce the neutron fluence. To understand the difference in results between the three configurations, it should be distinguished between these two effects. In this way it is evident that the latter effect (nuclear interactions) is larger in configuration 3 compared with configurations 1 and 2 which leads to a lower neutron fluence at the patient plane. 3.2. Neutron ambient dose equivalent at the patient plane Using the fluence to ambient dose equivalent conversion factors, taken from ICRP 74 (ICRP 1995), it is possible to evaluate the undesired neutron dose delivered at the patient plane in photon field configurations 1, 2 and 3, as shown in figure 10. The conversion factors (ICRP 1995) are strongly dependent on neutron energy, which must therefore be evaluated accurately in order to obtain an accurate dose estimate. In clinical treatments the amount of radiation delivered is specified in MU and thus it is important to link neutron ambient dose equivalent, expressed in terms of msv per Gy of photon dose, to delivered monitor units. A calibration process connects photon absorbed dose in a water phantom to delivered MU, with 100 MU usually being equivalent to 1 Gy at the depth of maximum dose (d max ), using a 10 10 cm 2 field. The photon dose at 3 cm depth in a water phantom has been calculated, using MCNP4B, for photon field configurations 1, 2 and 3, with the results summarized in table 2. Energy cut-offs of 5 kev for photons and 50 kev for

MC simulation of the photoneutron field in linac radiotherapy treatments 579 Figure 10. Undesired neutron ambient dose equivalent at the patient plane, calculated with MCNP4B-GN. Table 2. Photon dose at d max in a water phantom per electron incident on the target, calculated with MCNP4B-GN. MU calculated to obtain 1 Gy photon dose at d max. Field Jaws MLC Photon dose Uncertainty O R (cm 2 ) (cm 2 ) (Gy/e ) (%) D/D ref field MU 10 10 40 40 D ref field = 8.43 10 16 5.6 1 100 40 40 10 10 D = 8.16 10 16 5.9 0.968 103 10 10 Clinical D = 8.21 10 16 6.2 0.974 103 configuration electrons were used to speed up the calculations, without compromising on accuracy, and the calculated photon doses are expressed in units of Gy per source electron. Table 2 also gives the output factor, O R, defined as the ratio of photon dose for a particular field configuration to the photon dose obtained with reference field configuration 1. If 100 MU are required to deliver 1 Gy at 3 cm depth using reference field 1 then the number of MU required to obtain the same dose using field configurations 2, 3 is given by the equation (Dutreix and Svensson 1995): MU = D p (1) Ḋ R O R where D p is the prescribed dose: 1 Gy = 100 cgy, Ḋ R is the beam intensity (cgy MU 1 ) and O R is the output factor. The results are presented in table 2. Clearly the undesired photoneutron dose follows the number of MU and thus to evaluate this dose correctly (and hence assess the potential risks of radiation-induced secondary malignancies (Verellen and Vahhavere 1999)) it should be expressed in units µsv per MU. This is done in table 3 where the MC results displayed in figure 10 have been divided by the MU factors of table 2.

580 A Zanini et al Figure 11. Prostate treatment planning. Table 3. Neutron ambient dose equivalent at the patient plane normalized to 1 MU. Field Neutron ambient Neutron ambient Jaws MLC Distance from dose equivalent dose equivalent Uncertainty (cm 2 ) (cm 2 ) isocentre (cm) (msv Gy 1 ) (µsv MU 1 ) (%) 3 3.29 32.9 7.9 10 10 40 40 8 2.03 20.3 5.6 15 1.71 17.1 4.3 3 3.87 37.6 2.1 40 40 10 10 8 2.84 27.6 2.3 15 2.70 26.2 3.1 Clinical 3 2.35 22.8 2.7 10 10 configuration 8 1.16 11.3 3.6 15 1.08 10.5 3.6 3.3. Neutron dose in a realistic treatment procedure Usually tumours have to be treated using multiple-angle photon fields in order to deliver a large dose to the treatment volume without causing undue damage to the surrounding healthy tissue. For example, prostate radiotherapy could use one antero-posterior (AP) beam and two laterals (as shown in figure 11), giving a total dose of 76 Gy, which would be administered in 2 Gy sessions. The required MU per session would depend on the tumour depth and patient size, but the values given in table 4 are typical. Assuming that the photon collimator configuration would be similar to 3 above, MCNP4B- GN has been used to estimate the neutron dose equivalent received during this programme of treatment with the results summarized in table 5.

MC simulation of the photoneutron field in linac radiotherapy treatments 581 Table 4. Incident beam and photon dose per session in cgy. Beam MU cgy Antero-posterior 90 66.6 Lateral 140 66.7 Lateral 140 66.7 Table 5. Undesired neutron dose equivalent, produced by AP and lateral irradiations, at the patient plane at different distances from the isocentre, during a prostate treatment. Neutron ambient Total undesired Neutron ambient dose equivalent neutron dose Distance dose equivalent (1 session) (38 sessions) (cm) (µsv MU 1 ) MU (msv) (msv) Antero-posterior 3 22.8 90 2.05 77.9 8 11.3 90 1.02 38.8 15 10.5 90 0.95 36.1 Lateral 3 22.8 140 3.19 121.2 8 11.3 140 1.58 60.0 15 10.5 140 1.47 55.9 From table 5 it can be seen that much of the neutron contribution comes within the limits of the photon field, but outside these limits the decrease in neutron dose with distance from the isocentre is very small. In some situations, in conjunction with the MLC that defines the photon field edge, lead-alloy wedges are used. These wedges are situated inside the lateral photon field for a better distribution of the photon dose in the tumour area. The extra lead-alloy shielding reduces the efficiency of photon beam delivery, requiring more MU for a given treatment and additionally provides an extra source of photoneutrons. Therefore the undesired neutron dose could be higher than the results estimated in table 5. The next step of this research will be the simulation of a real treatment including lead-alloy wedges to give a precise assessment of the undesired neutron dose in critical organs. 4. Conclusions The Monte Carlo code MCNP4B-GN has been set up to evaluate secondary neutron doses which can be expected at bremsstrahlung radiotherapy facilities. The code incorporates a photoneutron event generator and considers neutron production and transport within a realistic model of the treatment apparatus and room. The calculations presented here show that neutron dose at the patient plane depends on the configuration of the heavy-metal collimators used to define the shape of the target volume. This configuration will affect not only the photoneutron production rate, but also the efficiency with which the photon dose is delivered to the target volume. If more MU are required to administer a specific photon dose to the target volume, then the accumulated neutron dose will rise and hence the neutron doses have been presented in units of µsv MU 1. The present calculations, made for an operational therapy facility, show that in the course of a typical treatment the neutron contribution is non-negligible and could represent a late

582 A Zanini et al risk for surrounding healthy tissues. Given that improvements to therapy procedures, such as increased beam energy and the use of MLC, have extended the life expectancy of patients (Schneider et al 2001) the potential late effects become more important. Thus, in line with the EURATOM 97 recommendations on patient radioprotection, therapy programmes must be tailored to minimize such secondary neutron doses and for this detailed, realistic calculations, as presented here, will be indispensable. Acknowledgments The authors wish to thank the IRCC radiotherapy staff, particularly Dr M Fabris, for their invaluable assistance. We acknowledge the support of the Max Lab (Lund, Sweden) research group and the UK Engineering and Physical Science Research Council. References Akkurt I et al 2003 Photoneutron yields from tungsten in the energy range of the giant dipole resonance Phys. Med. Biol. 48 3345 52 BCRU 1997 Advice on the implication of the conversion coefficients for external radiation published in ICRP Publication 74 and by ICRU Report 57 Radiat. Prot. Dosim. 74 195 8 Burn K W and Ongaro C 2002 Photoneutron production and dose evaluation in medical accelerators ENEA Report RT/2002/51/FIS (Bologna: ENEA) Dutreix A and Svensson H 1995 Monitor unit calculation and verification for therapy machines 3rd Biennial Meeting on Physics in Clinical Radiotherapy (Gardone Riviera, Italy) Gudowska I, Brahme A, Andreo P, Gudowski W and Kierkegaard J 1999 Calculation of absorbed dose and biological effectiveness from photonuclear reactions in a bremsstrahlung beam of end point 50 MeV Phys. Med. Biol. 44 2099 125 International Commission on Radiological Protection (ICRP) 1991 The 1990 1991 Recommendations of the International Commission on Radiological Protection ICRP Publication 60 (Ann. ICRP 21 (1 3)) (Stockholm: ICRP) International Commission on Radiological Protection (ICRP) 1995 Conversion coefficients for use in radiological protection against external radiation ICRP Publication 74 (Ann. ICRP 26 (3 4)) (Stockholm: ICRP) MCNP4B 1997 MCNP TM a general Monte Carlo N-particle transport code, version 4B LA-12625-M Manual ed J F Briesmeister (Los Alamos National Laboratory) Ongaro C, Burn K W, Zanini A, Nastasi U, Ottaviano G, Manfredotti C and Rodenas J 2000 Analysis of photoneutron spectra produced in medical accelerator Phys. Med. Biol. 45 L55 61 Ongaro C, Zanini A and Nastasi U 1999 Monte Carlo simulation of the photo-neutron production in the high-z components of radiotherapy linear accelerators Monte Carlo Methods Appl. 5 69 79 Schneider U, Lomax A and Lombriser N 2001 Comparative treatment planning using secondary cancer mortality calculations Phys. Medica XVII 97 9 Verellen D and Vahhavere F 1999 Risk assessment of radiation-induced malignancies based on whole-body equivalent dose estimates for IMRT treatment in the head and neck region Radiother. Oncol. 53 199 203