TYPE Original Research PUBLISHED 08 September 2022 DOI 10.3389/fonc.2022.903537 Simulation and experimental OPEN ACCESS verification of ambient neutron EDITED BY Valdir Carlos Colussi, University Hospitals Cleveland Medical doses in a pencil beam scanning Center, United States REVIEWED BY proton therapy room as a Chris Allgower, United States Department of Veterans Affairs, United States function of treatment Sunyoung Jang, Princeton Radiation Oncology Center, plan parameters United States *CORRESPONDENCE Olivier Van Hoey Olivier Van Hoey 1*, Liliana Stolarczyk 2,3,4, Jan Lillhök 5, ovhoey@sckcen.be Linda Eliasson 6, Natalia Mojzeszek 3, Malgorzata Liszka 3,4, SPECIALTY SECTION This article was submitted to Ali Alkhiat 4,7, Vladimir Mares 8, François Trompier 9, Radiation Oncology, Sebastian Trinkl 8,10, Immaculada Martı́nez-Rovira 11, a section of the journal Frontiers in Oncology Maite Romero-Expósito 11, Carles Domingo 11, Ondrej Ploc 12, RECEIVED 24 March 2022 Roger Harrison 13 and Pawel Olko 3 ACCEPTED 07 July 2022 1 Belgian Nuclear Research Center (SCK CEN), Institute for Environment, Health and Safety (EHS), Mol, PUBLISHED 08 September 2022 Belgium, 2Danish Centre for Particle Therapy, Aarhus University Hospital (AUH), Aarhus, Denmark, CITATION 3 Institute of Nuclear Physics, Polish Academy of Sciences, (IFJ PAN), Krakow, Poland, 4The Skandion Van Hoey O, Stolarczyk L, Lillhök J, Clinic, Uppsala, Sweden, 5Swedish Radiation Safety Authority, Solna, Sweden, 6Department of Physics, Eliasson L, Mojzeszek N, Liszka M, Royal Institute of Technology (KTH), Stockholm, Sweden, 7Department of Medical Radiation Physics Alkhiat A, Mares V, Trompier F, and Nuclear Medicine, Karolinska University Hospital, Stockholm, Sweden, 8Helmholtz Zentrum Trinkl S, Martı´nez-Rovira I, München, Institute of Radiation Medicine, Neuherberg, Germany, 9Institut de Radioprotection et de Romero-Expósito M, Domingo C, Sûreté Nucléaire (IRSN), PSE-Santé, Fontenay-aux-Roses, France, 10Federal Office for Radiation Ploc O, Harrison R and Olko P (2022) Protection, Neuherberg, Germany, 11Departament de Fı´sica, Universitat Autònoma de Barcelona, Simulation and experimental Bellaterra, Spain, 12Department of Radiation Dosimetry, Nuclear Physics Institute of the Czech Academy verification of ambient neutron doses of Sciences (CAS), Prague, Czechia, 13Faculty of Medical Sciences, University of Newcastle upon Tyne, in a pencil beam scanning proton Newcastle Upon Tyne, United Kingdom therapy room as a function of treatment plan parameters. Front. Oncol. 12:903537. doi: 10.3389/fonc.2022.903537 COPYRIGHT Out-of-field patient doses in proton therapy are dominated by neutrons. © 2022 Van Hoey, Stolarczyk, Lillhök, Currently, they are not taken into account by treatment planning systems. Eliasson, Mojzeszek, Liszka, Alkhiat, There is an increasing need to include out-of-field doses in the dose Mares, Trompier, Trinkl, Martı´nez-Rovira, Romero-Expósito, Domingo, Ploc, calculation, especially when treating children, pregnant patients, and patients Harrison and Olko. This is an open- with implants. In response to this demand, this work presents the first steps access article distributed under the towards a tool for the prediction of out-of-field neutron doses in pencil beam terms of the Creative Commons Attribution License (CC BY). The use, scanning proton therapy facilities. As a first step, a general Monte Carlo distribution or reproduction in other radiation transport model for simulation of out-of-field neutron doses was forums is permitted, provided the original author(s) and the copyright set up and successfully verified by comparison of simulated and measured owner(s) are credited and that the ambient neutron dose equivalent and neutron fluence energy spectra around a original publication in this journal is solid water phantom irradiated with a variation of different treatment plan cited, in accordance with accepted academic practice. No use, parameters. Simulations with the verified model enabled a detailed study of the distribution or reproduction is variation of the neutron ambient dose equivalent with field size, range, permitted which does not comply with these terms. modulation width, use of a range shifter, and position inside the treatment room. For future work, it is planned to use this verified model to simulate out- of-field neutron doses inside the phantom and to verify the simulation results by comparison with previous in-phantom measurement campaigns. Frontiers in Oncology 01 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 Eventually, these verified simulations will be used to build a library and a corresponding tool to allow assessment of out-of-field neutron doses at pencil beam scanning proton therapy facilities. KEYWORDS Proton therapy, Pencil beam scanned proton therapy, Neutron doses, Monte Carlo simulations, Out-of-field neutron doses in radiation therapy, Neutron measurements 1 Introduction doses in proton therapy as a function of position with respect to the isocenter and the beam direction and the most critical One of the biggest challenges in radiotherapy is to maximize treatment plan parameters such as field size, range, tumor damage, while sparing healthy tissues in order to modulation width, use of a range shifter, and air gap between minimize detrimental effects in these healthy tissues. With the range shifter and patient. This task focuses on active pencil proton therapy, the radiation energy can be deposited more beam scanning (PBS) proton therapy systems, as these systems locally in the tumor in comparison with photon therapy. This are now becoming standard in proton therapy (2) and limit the leads to improved healthy tissue sparing (1). Therefore, the use out-of-field neutron doses by avoiding significant creation of of proton therapy has been increasing rapidly over the last secondary neutrons in the proton delivery system. This task will decades with now over 100 active proton therapy facilities and lead to a better understanding of the relation between the out-of- over 250,000 patients treated worldwide (2). field neutron doses and the treatment plan parameters and allow However, despite the improved healthy tissue sparing, there medical physicists to evaluate and, if necessary, adapt the is still some dose deposited in healthy tissues due to secondary treatment plan also with respect to the out-of-field neutron and scattered radiation. The out-of-field doses in proton therapy doses and associated detrimental effects. In the end, this will are dominated close to the target by secondary protons and contribute to improving the patient’s life expectancy and further away from the target by secondary neutrons and gamma life quality. radiation created by interactions of protons with the beamline, This work presents the first steps that were performed within the patient, and the room. These out-of-field doses can lead to this task. An MC radiation transport simulation model was set detrimental effects in healthy tissues and should be considered up to simulate secondary radiation production and transport and possibly minimized during the treatment planning. during patient treatment in PBS proton therapy facilities. For Several studies have already characterized the out-of-field verification of the MC model, the neutron doses simulated neutron doses in proton therapy for some specific cases using outside an irradiated phantom were compared with ambient Monte Carlo (MC) radiation transport simulations, neutron monitor measurements at different positions close to measurements, or analytical models (3, 4). However, it is not the phantom for varying treatment plan parameters at two PBS straightforward to compare neutron doses in the literature and proton therapy facilities. The goals of this study were to verify to estimate the neutron dose for a specific patient based on these the MC model and to investigate the variation of the ambient studies due to the strong dependence of the neutron dose on the neutron doses as a function of position, field size, range, treatment plan parameters. The typical normalization of the out- modulation width, use of a range shifter, and air gap between of-field neutron doses to the absorbed dose in the target or the the range shifter and phantom. product of absorbed dose in the target and the treatment volume is not sufficient to allow direct comparison of out-of-field doses from different studies. Moreover, current treatment planning 2 Materials and methods systems (TPS) do not take into account out-of-field neutron doses. However, there is an increasing need to include out-of- 2.1 Measurements field neutron doses in the TPS, especially when treating children, pregnant patients, and patients with implants such as 2.1.1 Proton therapy facilities pacemakers or hearing implants. For verification of the MC simulations, measurements were In response to this demand, within EURADOS WG9, a performed at two PBS proton therapy facilities. A first dedicated task was set up. This task has the final aim to provide exploratory measurement campaign was performed at the an easy-to-use tool to quickly assess the out-of-field neutron Bronowice Cyclotron Center (CCB) Institute of Nuclear Frontiers in Oncology 02 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 Physics (IFJ PAN) in Krakow (Poland) in May 2017. Based on radiation protection quantity effective dose. Due to limitations in the experience from this measurement campaign, a second more experimental time and ambient neutron monitor availability, it extensive measurement campaign was performed at the was not possible to use the same type of ambient neutron monitor Skandion Clinic in Uppsala (Sweden) in July 2019. Both at all six positions. However, the specific aspects and uncertainties facilities are equipped with gantries with a dedicated scanning of each monitor were taken into account in the data analysis. nozzle (IBA Proton Therapy System – Proteus 235). The range Therefore, different types of ambient neutron monitors from the shifter at CCB is fixed at the nozzle, whereas at Skandion, it can different institutes participating in the measurement campaigns be moved on the snout toward the patient. were used. However, each ambient neutron monitor was kept at a fixed position throughout the whole measurement campaign. The 2.1.2 Experimental setup institute and type of the ambient neutron monitors used at the six During both measurement campaigns, rectangular target measurement positions in Skandion and CCB are listed in Table 2. volumes were delivered from the side with the nozzle at 270° On the second and third lines in Table 2, also the manufacturer to a 30 cm × 30 cm × 60 cm solid water phantom placed on the and the calibration date and source are specified for each monitor. treatment table. The isocenter was positioned at 15 cm depth in All calibrations were still valid at the time of the measurement the phantom and at 15 cm from top, bottom, and the three campaigns according to the calibration procedures of the closest side faces of the phantom. The setup was similar to that respective institutes. described in (5). Treatment plans were prepared using Varian A very important characteristic of ambient neutron Eclipse treatment planning systems (version 13.6 and 15.6 in monitors is their energy response in terms of H*(10). Ideally, CCB IFJ PAN and Skandion, respectively). Plan variables the H*(10) energy response should be close to unity for all included field size, range, Spread Out Bragg Peak (SOBP) possible neutron energies. However, in practice, no existing modulation width, use of a range shifter, and air gap between ambient neutron monitor has a perfect energy response for the the range shifter and solid water phantom. The dose delivered to wide range of possible neutron energies. This has to be taken the center of the SOBP for each irradiation was 20 Gy in CCB IFJ into account when analyzing the measurement results and PAN and 5 Gy in Skandion. An overview of the covered plans for comparing them with the MC simulations. The different both measurement campaigns is shown in Table 1. Data for ambient neutron monitors are described in detail below. The irradiations with the range shifter are presented in this work only H*(10) energy response functions are compared in Figure 2. for the Skandion proton therapy facility. To obtain the Where necessary, they were normalized taking into account the prescribed range and modulation width, for each plan, a set of energy spectrum of 252Cf or Am-Be neutrons used during the proton energies was used. Minimal and maximal proton energies calibration. For HAWK, these response data are obtained from are also given in Table 1. measurements, while for the other monitors, the response data Six ambient neutron monitors were positioned around the are obtained from simulations. solid water phantom at six fixed positions labeled A–F. A Neutron energy spectra at proton therapy facilities typically schematic representation and pictures of the setup and the exhibit peaks for thermal, evaporation, and high-energy measurement positions at both facilities are shown in Figure 1. neutrons (11). The energy spectra of 252Cf or Am-Be neutrons fall in the same energy range as the evaporation neutrons. The 2.1.3 Ambient neutron monitors energy ranges of these peaks are indicated in Figure 2. The MC simulations were verified by measurements with Figure 2 shows that the energy response is close to unity for ambient neutron monitors during the measurement campaigns at all monitors in the evaporation neutron energy range. This is Skandion and CCB. Ambient neutron monitors measure the expected because they are calibrated with 252Cf or Am-Be neutron dose in terms of the quantity ambient dose equivalent neutrons in the same energy range. In the other energy ranges H*(10). This quantity, as defined in (6), is a measurable there are significant deviations from unity and also significant operational quantity that provides a conservative estimate of the differences between the different monitors. TABLE 1 Overview of covered treatment plans during the measurement campaigns at the CCB IFJ PAN and Skandion proton therapy facilities with PBS. Center Range shifter (RS) Air gap RS- Field size Proton energies Range SOBP modulation phantom [cm] [cm2] [MeV] [cm] width [cm] CCB IFJ PAN (Krakow, – – 25–400 Min: 74–148 15–25 10–20 Poland) Max: 146–192 Skandion Clinic 3.1 cm WET Lexan on 5.5–23 9–625 Min: 65–103 8–25 3–25 (Uppsala, Sweden) movable snout Max: 103–212 Frontiers in Oncology 03 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 1 Schematic representation of the setup and the measurement positions (left), picture of the measurements at Skandion (right top), and picture of the measurements at CCB (right bottom). One can see in Figure 2 that in the thermal energy range, the available in the epithermal energy range. However, neutrons in ambient neutron monitors exhibit an under-response between a the thermal and epithermal energy range are not expected to factor of two and five. It can also be seen in Figure 2 that contribute more than a few percent to the neutron H*(10) in conventional ambient neutron monitors such as LB 6411, proton therapy treatment rooms (12). Therefore, the under- Wendi-II, NM2B-495Pb, and NM2B-458 exhibit a typical response in the thermal neutron range and the over-response over-response in the epithermal energy range related to the and under-response of respectively conventional ambient very strong thermalization of epithermal neutrons in their neutron monitors and TEPCs in the epithermal neutron range moderator sphere. On the other hand, Tissue Equivalent are not expected to affect the measurements significantly. Proportional Counters (TEPCs) such as Sievert have an under- Finally, one can observe in Figure 2 that the H*(10) energy response in the epithermal energy range due to lower response of LB 6411 and NM2B-458 drops to zero above 10–20 backscattering in the gas volume in comparison with the MeV due to their very limited sensitivity for high-energy ICRU sphere in the definition of H*(10). A similar under- neutrons. The sensitivity is increased for Wendi-II and NM2B- response is expected for HAWK, but no response data are 495Pb by an additional metallic shield embedded in the FIGURE 2 Plot with the simulated (lines) or measured (data points) H*(10) energy response functions of the different ambient neutron monitors used during the measurement campaigns at Skandion and CCB [LB 6411 (7), Wendi-II (8), Sievert [this work], NM2B-495Pb and NM2B-458 (9), HAWK (10)]. Frontiers in Oncology 04 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 TABLE 2 Ambient neutron monitors used at the different positions at Skandion and CCB. Position Skandion July 2019 CCB May 2017 A SSM Sievert CCB Wendi-II SSM, Stockholm, Sweden Thermo Scientific, Waltham, USA Calibration: 09-2018, 252Cf and Am-Be Calibration: 06-2016, 252Cf and Pu-Be B SCK CEN Wendi-II SCK CEN Wendi-II Thermo Scientific, Waltham, USA Thermo Scientific, Waltham, USA Calibration: 05-2018, 252Cf Calibration: 11-2015, 252Cf C SSM Sievert HMGU NM2B-495Pb SSM, Stockholm, Sweden NE Technology Ltd., Benham, UK Calibration: 09-2018, 252Cf and Am-Be Calibration: 03-2017, Am-Be D Skandion LB 6411 UAB LB 6411 Berthold, Bad Wildbad, Germany Berthold, Bad Wildbad, Germany Calibration: 08-2013, Am-Be Calibration: 06-2010, Am-Be E IRSN HAWK HMGU NM2B-458 Far West Technologies, Puyallup, USA NE Technology Ltd., Benham, UK Calibration: 06-2019, Am-Be Calibration: 03-2017, Am-Be F SSM LB 6411 NPI LB 6411 Berthold, Bad Wildbad, Germany Berthold, Bad Wildbad, Germany Calibration: 09-2018, 252Cf and Am-Be Calibration: 10-2014, Am-Be SSM, Swedish Radiation Safety Authority, Stockholm, Sweden. SCK CEN, Belgian Nuclear Research Center SCK CEN, Mol, Belgium. Skandion, Skandionkliniken, Uppsala, Sweden. IRSN, Institut de Radioprotection et de Sû reté Nuclé aire, Fontenay-aux-Roses, France. CCB, Bronowice Cyclotron Center IFJ PAN, Krakow Poland. HMGU, Helmholtz Zentrum Munchen, Munich, Germany. UAB, Universitat Autònoma de Barcelona, Departament de Fı́sica, Cerdanyola del Vallès, Spain. NPI, CAS, Nuclear Physics Institute, Prague, Czech Republic. moderator. However, even for these monitors, there is still a where q is the mean collected charge during the time interval, limited under-response for high-energy neutrons. TEPCs such W/e is the average energy required to create an ion pair, M is the as Sievert and HAWK are intrinsically sensitive for high-energy gas multiplication factor, and mdet the mass of the detector neutrons. However, for Sievert, one can see that there is a limited gas mass over-response for high-energy neutrons. As high-energy The dose-mean lineal energy is calculated using the variance neutrons can deliver a large fraction of the neutron dose in method proton therapy rooms (12), the under-response and over- mdet VD,rel  response of the ambient neutron monitors for high-energy yD = l Ddet neutrons is expected to affect the measurement results significantly. This will be discussed in detail in Sections 3.1.1.4 where VD,rel is the relative variance in the absorbed dose during and 3.2.2. repeated charge integrations and l¯ is the mean chord length of the simulated tissue volume 2.1.3.1 Sievert The dose equivalent H* is in turn determined by The two Sievert instruments used by SSM are TEPCs, made  det ða + byD Þ H* = D of A-150 plastic and filled with a propane-based tissue- equivalent gas held at 1.37 kPa pressure, simulating a tissue where the constants a = 0.88 and b = 0.09 μm/keV are chosen for volume with a mean chord length of 1.88 μm. The sensitive the typical high-energy neutrons present in the proton therapy volume is 1.207 dm3, with 11.5 cm for both height and diameter. rooms (13). Thus, by measuring the dose-mean lineal energy The electric charge is measured using a capacitor feedback using the TEPC, the dose equivalent in a mixed radiation field electrometer, where the voltage over a 1-nF capacitor is can be determined. measured over a charge collection time of 0.1 s (13). The absorbed dose fractions due to low and high LET The average absorbed dose during a charge collection time is radiation are calculated from the measuredyD value and the given by dose-mean lineal energies of photon and neutron components, that is,  det = qW=e D   M mdet yD = yD,g dg + yD,n dn = yD,g dg + yD,n 1 − dg Frontiers in Oncology 05 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 The dose-mean lineal energies for photonsyD,g = 1:4 keV=μm the dose equivalent Hhigh * is used in this work as an and neutrons yD,n = 96 keV=μm are calculated from MC- approximation of the neutron H*(10). simulated relative contribution in the mixed field. From the For the H*(10) energy response of the HAWK, only a limited relative dose contributions, the dose equivalent for photons and number of data points in the energy range from 0.5 to 60 MeV are neutrons, Hg* and Hn* are estimated from available from (10). These are also shown in Figure 2. It can be seen Hn* = dn Ddet (a + byD,n )and that the response is very close to unity in this energy range. Similar to   the Sievert and other TEPCs, also for the HAWK, an under-response Hg* = (1 − dn )Ddet a + byD,g for energies below 0.5 MeV can be expected. For high-energy neutrons, the response is expected to be close to unity as well. Using the relative dose fractions, a value of 28 eV was obtained for W/e. In this work, single spot measurements at proton beam 2.1.3.3 LB 6411 energies of 70, 146, and 212 MeV were performed to calculate yD The LB 6411 ambient neutron monitor from Berthold values for different energies and positions. To determine the Technologies used by SSM, UAB, and NPI consists of a 25- dose equivalent for the scanned irradiations, the maximum cm-diameter polyethylene moderator sphere with internal Cd proton energy was used to choose which yD value to use. NoyD absorbers and perforations that surrounds a cylindrical 3He measurements were performed with a range shifter. Therefore, proportional counter (7). The neutron sensitivity is around 3 no H*(10) measurements for the irradiations with the range counts per nSv of neutron H*(10). It has excellent photon shifter are presented for positions A and C at Skandion where rejection capabilities with less than 30 μSv/h of photon the Sievert was used at those positions. response in a 10 mSv/h photon radiation field. The Sievert H*(10) energy response plotted in Figure 2 was This ambient neutron monitor is designed to measure calculated by means of MC simulations. The response is fairly thermal to 20 MeV neutrons and it is known to have a close to unity over the whole energy range, except for an under- strongly decreasing sensitivity to neutrons above 20 MeV. This response for epithermal neutrons and a slight over-response for is clearly reflected in the H*(10) energy response plotted in high-energy neutrons. Figure 2, as obtained from MC simulations (7). The response is fairly close to unity over the whole energy range, except for an over-response in the epithermal energy range and a strong 2.1.3.2 HAWK under-response for neutrons with energies above about 20 The HAWK environmental Monitoring System FW-AD1 MeV for which the LB 6411 is almost insensitive. from Far West Technology Inc. used by IRSN is a microdosimetric single-event TEPC-system. The detector is spherical with 127 mm diameter (Benjamin type) and filled 2.1.3.4 Wendi-II with pure propane gas at 933 Pa to simulate an energy The Wendi-II ambient neutron monitor from Thermo deposition in 2 μm biological site and a mean chord length of Scientific used by SCK CEN and CCB is an extended-range 1.33 μm (14). HAWK measures the energy deposition spectrum ambient neutron monitor designed by Olsher et al. (16). It from particles correlated to a single initial particle event on a consists of a cylindrical polyethylene moderator with an inner lineal energy scale calibrated using a proton edge calibration. tungsten shell that surrounds a cylindrical 3He proportional The absorbed dose distribution in lineal energy d(y) and the low- counter. The neutron sensitivity is around 3 counts per nSv of and high-LET components are defined as the contributions neutron H*(10). It has excellent photon rejection capabilities below and above 10 keV/μm respectively. From this with less than 5 μSv/h of photon response in a 100 mSv/h photon distribution, the dose equivalent is calculated according to radiation field. The tungsten shell embedded in the polyethylene moderator * + H* H * = Hlow greatly enhances the Wendi-II response to high-energy high Z Z neutrons, extending the measurement range to about 5 GeV 10 1024 = Nlow Qð yÞd ð yÞdy + Nhigh Qð yÞdð yÞdy and thus well beyond the maximum neutron energy of about 200 0:5 10 MeV encountered in proton therapy facilities. This is clearly Here, Nlow and Nhigh are the low-LET and high-LET H*(10)- reflected in the H*(10) energy response plotted in Figure 2, as correction calibration factors from 137Cs and 60Co photon and obtained from MCNPX 2.7 simulations (8). The response is Am-Be neutron fields (10, 15). Nlow aims to compensate for the fairly close to unity over the whole energy range, except for an relatively high value of the electronic threshold, making it over-response for epithermal neutrons. impossible to measure events below 0.5 keV/μm. Nhigh aims to compensate for the uncorrected y value for the proton edge set 2.1.3.5 NM2B-458 and NM2B-495Pb by the manufacturer for the y scale calibration. Nlow is equal to A conventional NM2B‐458 and an extended‐range NM2B‐ 1.1 ± 0.02 and Nhigh to 0.8 ± 0.09. The high-LET component of 495Pb ambient neutron monitor were used by HMGU. These Frontiers in Oncology 06 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 monitors are based on the Andersson-Braun (AB) model and * a mesh tally inside the phantom to tally the SOBP manufactured by NE Technology Ltd. They consist of a delivered by the proton beam (red); cylindrical BF3 proportional counters of 3.1 cm diameter and * the 3.11-cm-thick and 30 cm by 30 cm cross section range 7.2 cm active length surrounded by an inner 1.7-cm-thick shifter made of Lexan with a density of 1.2 g/cm³ moderating polyethylene layer, a 0.6-cm-thick boron‐doped positioned with the appropriate air gap with respect to synthetic rubber absorber, and an outer 6.9-cm-thick the phantom edge for the cases in which a range shifter polyethylene moderator. The NM2B‐495Pb ambient neutron was used (yellow); and monitor additionally has a 1 cm thick lead shell surrounding the * six 20-cm-diameter spherical air cells at the detector boron rubber to increase the response to high-energy neutrons. positions for tallying the neutron fluence energy The H*(10) energy responses of both ambient neutron spectrum and neutron H*(10) (blue). monitors were calculated by means of MC simulations in the energy range from thermal to 10 GeV (9) and are shown in The advantage of this model is that it is generally valid for Figure 2. The response is fairly close to unity over the whole any PBS proton therapy facility after adaptation of the range energy range up to about 10 MeV for both ambient neutron shifter dimensions. The disadvantage is that without inclusion of monitors, except for an over-response for epithermal neutrons. the walls, floor, ceiling, patient table, gantry cylinders, gantry For higher-energy neutrons, only the NM2B-495Pb has a cone, counterweight, and other components in the room, the response fairly close to unity due to the additional lead shell. neutrons created inside these components are not taken into For the NM2B-458, the response for high-energy neutrons drops account. The scattered thermal and epithermal neutrons are not very rapidly above about 10 MeV. expected to contribute more than a few percent to the total neutron H*(10) (12). However, evaporation neutrons that are 2.2 Simulations partly created in iron-rich components such as the gantry cylinders, the gantry cone, and the counterweight (18) can All the simulations in this work were performed with the contribute up to about 50% of the total neutron H*(10) (19). MC radiation transport code MCNP6.2 (17). As the goal of this work is to develop a tool to predict out-of- The first important input in these simulations is the field neutron doses in any PBS proton therapy facility and to get geometrical model. Two different geometrical models were a better understanding of the trends of the out-of-field neutron implemented in this work. Firstly, a model was developed that doses as a function of treatment plan parameters, development does not take into account the specifics of the room. This model of detailed room models of the Skandion and CCB proton is shown on the left in Figure 3. It consists of therapy facilities was beyond the scope of this work. However, in order to better understand the potential sources of deviations * the 60-cm-long and 30 cm by 30 cm cross section solid between simulations and measurements, it was decided to also water phantom made of white polystyrene type RW3 develop a second geometrical model for the Skandion facility with 2% by weight TiO2 and a density of 1.03 g/cm³ with a simplified representation of the room. This model is (green); shown on the right in Figure 3. The top picture shows a view FIGURE 3 Geometrical models implemented in MCNP6.2. (Left) Model without room: solid water phantom (green), SOBP mesh tally (red), range shifter (yellow), and detectors (blue). (Right) Model with room with a view through the roof inside the room (top) and a view of the outside of the room (bottom): solid water phantom (green), positioning uncertainty mesh tally (red), range shifter (yellow), detectors (blue), table (orange), and walls, ceiling, and floor (purple). Frontiers in Oncology 07 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 inside the room through the roof, while the bottom picture the height of the detectors to obtain an estimation of the positioning shows an outside view of the room. In addition to the uncertainty of the measurements. The simulation results are all components implemented in the first model, this second expressed per simulation particle. However, as also the SOBP model also contains absorbed dose profile is simulated, all simulation results can be divided by the absorbed dose in the center of the SOBP per * the 6-cm-thick and 220 cm by 52 cm table made of Kevlar simulation particle in order to obtain the results expressed per with a density of 1.44 g/cm³ (orange); unit of absorbed dose in the center of the SOBP. This allows direct * walls, ceiling, and floor made of 1-m-thick ordinary NIST quantitative comparison between the simulations concrete (20) with a density of 2.3 g/cm³ (purple); and and measurements. * a mesh tally for assessing the positioning uncertainty of the measurements (red). 3 Results and discussion The 1-m-thick concrete everywhere is an approximation and might lead to an overestimation of the scattered neutrons in the 3.1 Uncertainties gantry area. Also, the iron-rich components of the gantry are not For clarity of the plots presented here, uncertainty bars were taken into account. This might lead to an underestimation of the not always added to the measurement and simulation results and evaporation neutrons that are, to a great extent, created in these their ratios. However, the order of magnitude of the components (18). uncertainties and the different contributions to the The second meaningful input in the simulations is the uncertainties are discussed in detail in this subsection. All definition of the radiation source. The rectangular proton uncertainties presented here are expressed at the k = 1 level. fields were modeled as a monodirectional collimated uniform square proton beam with size corresponding to the field size. The proton energy is sampled from the superposition of Gaussian 3.1.1 Measurements pffiffiffi 3.1.1.1 Positioning uncertainty energy distributions with 2s = 1:5 % and energies and weights obtained from the layer information in the *.pld file from The positioning uncertainty is estimated by assuming that the TPS. the positioning of the detectors was done with a precision of The neutron reaction cross sections were taken from the about 5 cm. The associated uncertainty on the measured H*(10) ENDF/B-VIII.0 database (21) for a temperature of 0.02585 eV. was estimated based on the simulated type 1 dose mesh tally with Both protons and neutrons were tracked with a high-energy ICRP 74 fluence to H*(10) conversion coefficients in the cutoff at 240 MeV, well above the maximum proton energy of horizontal plane at the height of the detectors. The simulated 212 MeV. The Bertini Dresner intranuclear cascade and H*(10) on the mesh tally was averaged over the volume of a evaporation model was selected because it gave reliable results detector, shifting the detector positions 5 cm from the reference in previous simulations of neutron doses in proton therapy positions in the positive and negative direction along the x, y, facilities (4, 22). Furthermore, a limited sensitivity study showed and z axes. The positioning uncertainty was then assessed less than 10% difference in simulated neutron H*(10) values assuming a uniform distribution between the lowest and when using the default cross sections and physics models highest dose values obtained for each detector position. The in MCNP6.2. obtained uncertainties averaged over all treatment plan Finally, it is necessary to define the appropriate tallies in the parameters vary between 2.6% for position A and 5.0% for simulations in order to obtain the desired quantities. A type 3 position B with a global average of 4%. As expected, a higher volumetric energy deposition mesh tally was defined in the target positioning uncertainty is found for the detector positions closer volume of the solid water phantom for assessing the SOBP absorbed to the isocenter due to the higher dose gradient. dose profile. In this way, it could be checked whether the proton beam was modeled in a sufficiently realistic manner. F4 fluence 3.1.1.2 Calibration and instrument specific uncertainty energy spectrum and F4 H*(10) tallies were defined in the detector All ambient neutron monitors were calibrated using 252Cf or cells. The energy bin width in lethargy was fixed at ln( EEii−1 ) = 0:26. Am-Be neutron sources. The calibration and instrument-specific This allowed direct comparison with the neutron H*(10) values uncertainty is estimated to be 5% for Sievert, 6% for HAWK, and measured by the ambient neutron monitors in this work and the about 2% for the other ambient neutron monitors. neutron energy spectra measured with Bonner spheres during a previous measurement campaign at a similar PBS proton therapy 3.1.1.3 Statistical uncertainty facility in Trento (Italy) (11). A type 1 dose mesh tally with ICRP 74 The statistical uncertainty is the uncertainty on the fluence to H*(10) conversion coefficients using the TMESH DOSE measurement itself related to Poisson counting statistics. The keyword with option ic = 40 was defined in the horizontal plane at relative statistical uncertainty was calculated as the inverse Frontiers in Oncology 08 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 square root of the number of counts obtained during the 3.1.1.5 Target dose uncertainty measurement. This uncertainty ranges between 6% and 18% Finally, the uncertainty on the delivered absorbed dose in the with an average of 10% for Sievert at position A, between 9% and target volume also contributes to the measurement uncertainty as 25% with an average of 20% for Sievert at position C, between both measurements and simulations are normalized to the 1% and 3% for HAWK and well below 1% for the other ambient absorbed dose in the target volume. This uncertainty is neutron monitors in most cases with exceptions up to 2% for estimated to be about 2.5%. This uncertainty also includes the very low doses. day-to-day variations of less than 0.5% based on the daily QA measurements. 3.1.1.4 Energy response uncertainty The energy response uncertainty is related to the imperfect 3.1.1.6 Combined uncertainty H*(10) energy response of the ambient neutron monitors. The From the discussion above, it is clear that the measurement expected responses of the different ambient neutron monitors at uncertainty varies significantly. It depends on the type of ambient their measurement positions were estimated by convoluting the neutron monitor, the measurement position, and the treatment neutron H*(10) energy spectra simulated with the model plan parameters. Combining all the above uncertainties, it can be without the room specifications at their measurement position estimated that the combined measurement uncertainty is typically with the monitor H*(10) energy response. The minimum, in the range between 15% and 30%. maximum, and average H*(10) responses for the different treatment plan parameters for each monitor are shown 3.1.2 Simulations in Table 3. The number of particles in the MCNP6.2 simulations was No data are shown for HAWK because the H*(10) energy taken sufficiently high to keep the statistical uncertainties on the response was only available for a limited energy range. However, as simulated H*(10) values below 1% and on the simulated fluence discussed in Section 2.1.3.2, its response is expected to be close to energy spectra below 5% for all energy bins contributing unity for the measurements in this work. The NM2B-495Pb with significantly to the total fluence. However, the statistical extended energy range seems to perform worse than the NM2B-458 uncertainties of MC simulations are only a minor component without extended energy range. However, it has to be taken into of the total uncertainty. The total uncertainty is dominated by account that the NM2B-495Pb was used at position C, which is in uncertainties in the reaction cross sections, uncertainties in the the forward direction where high-energy neutrons are expected to physics models and simplifications or inaccuracies in the model contribute significantly to the dose, whereas the NM2B-458 was used geometry. Assessment of these uncertainties is not at position E in the backward direction where virtually no high- straightforward and was considered beyond the scope of this energy neutrons are expected. work. It just has to be kept in mind when comparing the It can be seen that significant overestimations up to 47% and measurements and simulations that the simulation results significant underestimations up to 67% are possible. Hence, the come with a significant uncertainty as well. energy response will often be the biggest source of measurement uncertainty. Sievert shows a systematic overestimation, while all other ambient neutron monitors exhibit a systematic 3.2 Verification of the simulations underestimation. This is well in line with the over-response and under-response for high-energy neutrons for respectively 3.2.1 SOBP depth profile the Sievert and the other ambient neutron monitors as shown in First, it was checked whether the proton beam was modeled Figure 2 and discussed in Section 2.1.3. in a sufficiently realistic way. For this, the simulated SOBP TABLE 3 Minimum, maximum, and average H*(10) responses of the different ambient neutron monitors at the different positions A-F as estimated from the convolution of the neutron H*(10) energy spectra simulated with the model without room specifications and the simulated H*(10) energy responses. Position Skandion CCB Detector Min Max Average Detector Min Max Average A SSM Sievert 1.34 1.42 1.38 IFJ Wendi-II 0.78 0.86 0.81 B SCK CEN Wendi-II 0.72 0.97 0.85 SCK CEN Wendi-II 0.78 0.99 0.88 C SSM Sievert 1.40 1.47 1.45 HMGU NM2B-495Pb 0.59 0.66 0.62 D Skandion LB 6411 0.33 0.61 0.46 UAB LB 6411 0.35 0.53 0.44 E IRSN HAWK – – – HMGU NM2B-458 0.81 0.90 0.86 F SSM LB 6411 0.58 0.72 0.66 NPI LB 6411 0.63 0.73 0.69 Frontiers in Oncology 09 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 profiles were plotted and compared with the range and (10) is compared with the simulated H*(10) from the simulations modulation width of the corresponding treatment plan. The with the general geometrical model not taking into account the shape of the SOBP profiles agreed very well with the expected specifics of the room. In order to evaluate the uncertainty related shape based on the range and modulation width for all cases. to disregarding the specifics of the room, in the second case, the Three representative cases are shown in Figure 4. In this plot, the measured H*(10) is compared with the simulated H*(10) from the SOBP profiles are normalized to the average of the plateau of the simulations with the simplified Skandion room model. For SOBP. Hence, it could be concluded that the proton beam was assessing the uncertainty related to the imperfect ambient modeled in a sufficiently realistic way. neutron monitor H*(10) energy response, in the third case, the measured H*(10) is compared with the H*(10) obtained by 3.2.2 Neutron ambient dose equivalent convolution of the simulated neutron H*(10) energy spectra Further verification of the simulation model was performed by from the simulations without the specifics of the room with the comparison of the neutron ambient dose equivalent H*(10) from H*(10) energy response of the appropriate ambient neutron the simulations with the measured H*(10) from the ambient monitor. For position E, the third case is not applicable, because neutron monitors. The first measurement campaign was the H*(10) energy response of the HAWK used at this position is performed at CCB for a limited set of treatment plan parameters only available for a limited energy range. Table 4 gives an overview in May 2017. Later, in July 2019, a more extensive measurement of the average ratio of simulated over measured H*(10) per campaign was performed at Skandion. The comparison between position and averaged over all positions for the three cases. The simulations and measurements will be made here in detail for the separate ratios of simulated over measured H*(10) for all Skandion measurement campaign as it provides the most extensive irradiations separately are plotted in Figure 5 (position A, B, data set. Furthermore, an approximate model of the Skandion and C) and in Figure 6 (position D, E, and F) as a function of the treatment room was implemented in the simulations as explained in irradiation number as specified in Table A.1 in the appendix. The Section 2.2. In Section 3.3, the variation of the H*(10) with position ratios for the simulations without specifics of the room are shown and treatment plan parameters, and the measurements at Skandion as red crosses, the ratios for the simulations with specifics of the and CCB are also compared with each other. An overview of all the room are shown as blue circles, and the ratios for the simulations irradiations and their treatment plan parameters for the without specifics of the room and correction for the imperfect measurement campaigns performed at Skandion and CCB is ambient neutron monitor H*(10) energy response are shown as shown in Tables A.1 and A.2 in the appendix. Each of the listed green triangles. A green line indicating a ratio of one and two red irradiations was performed only once due to beam time limitations lines indicating a factor of two over- or under-response are added and expected stability of the irradiations. to guide the eye. Error bars are not added here for clarity of the The comparison between simulated and measured H*(10) is plots. However, a more detailed comparison with error bars is made for three different cases. In the first case, the measured H* presented in Sections 3.3.2–3.3.5. FIGURE 4 Plot of the simulated SOBP profiles for three different cases: range, 10 cm; modulation width, 3 cm (green full line); range, 15 cm; modulation width, 5 cm (blue dashed line); range, 25 cm; modulation width, 10 cm (red dotted line). Frontiers in Oncology 10 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 5 Plots with the ratio of simulated over measured H*(10) for positions (A–C) for the Skandion measurement campaign. Three cases are shown: simulations without room (red crosses), simulations with room (blue circles) and simulations without room and correction for the imperfect ambient neutron monitor H*(10) energy response (green triangles). A green line indicating a ratio of one and two red lines indicating a factor of two over- or under-response are added to guide the eye. FIGURE 6 Plots with the ratio of simulated over measured H*(10) for positions (D–F) for the Skandion measurement campaign. Three cases are shown: simulations without room (red crosses), simulations with room (blue circles), and simulations without room and correction for the imperfect ambient neutron monitor H*(10) energy response (green triangles). A green line indicating a ratio of one and two red lines indicating a factor of two over- or under-response are added to guide the eye. Frontiers in Oncology 11 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 One can see in Figures 5, 6 that even without the room in the overestimation of the scattered neutron dose contribution by simulation model and without energy correction, most of using 1-m-thick concrete everywhere in the gantry area in the the simulation data are already within a factor of two from the simulation model. Furthermore, there are also uncertainties in measurements. Only for position D and for one irradiation at the concrete composition and concrete hydrogen content that position E are there stronger underestimations. From the results can both strongly affect the scattered neutron contribution. in Table 4 and Figures 5, 6, one can also see that introduction of Correction for the imperfect energy response of the ambient the room in the simulation model significantly increases the neutron monitors leads to a decrease of the measurements for simulated H*(10) values. This systematic increase for all positions A and C and an increase for position B as expected irradiations is caused by the additional scattered neutrons from the responses tabulated in Table 3. Correction for the from the room contributing to the neutron dose. For positions imperfect energy response worsens the agreement between A, C, E, and F, the increase is about 60%. For position D, a much measurements and simulations. This can be caused by higher increase with almost a factor of three is observed. This is uncertainties in the neutron monitor energy response caused by the fact that position D is the most distant position functions, uncertainties in the simulated neutron fluence from the gantry area perpendicular to the beam direction, where energy spectra due to absence of the room and other scattered neutrons are expected to contribute the most. There important components influencing the energy spectra in the might also be an important dose contribution from neutrons simulations, and the strong sensitivity of the neutron monitor created inside the bending magnet of the gantry, which are not response on the neutron energy. These observed effects of taken into account in the simulations. These additional scattered scattered neutrons from the room and the imperfect energy neutrons are probably the reason for the underestimation of the response of the ambient neutron monitor can also account for neutron dose at position D in the simulations. On the other the limited deviations found between the measurements and hand, for position B, the increase is limited to only 12%. This is the simulations. also expected because position B is the position closest to the F o r p o s i t i o n D , o n e c an ob s e r v e a s y s t e m a t i c isocenter at only 1 m in the direction of the beam, where underestimation of 56% for the simulations without the scattered neutrons are not expected to contribute significantly. specifics of the room in comparison with the measurements. From the results in Table 4 and Figures 5 and 6, one can also This underestimation could be expected because position D is see that when correcting for the imperfect energy response of the 2.25 m away from the isocenter in a direction perpendicular to ambient neutron monitors, the ratio of simulations over the proton beam. There, neutrons scattered by the room are measurements systematically increases with about 40% for the expected to contribute significantly to the neutron dose. This is Sievert measurements at positions A and C, while it confirmed by the fact that the simulated H*(10) values for systematically decreases about 15%, 50%, and 35% for position D for the simulations with the specifics of the room respectively Wendi-II at position B and LB 6411 at positions are significantly higher and on average 23% above the measured D and F. This is related to the imperfect energy response of the H*(10) values. This overestimation can be related to ambient monitors for high-energy neutrons as discussed in uncertainties in the concrete composition and concrete Sections 2.1.3 and 3.1.1.4. hydrogen content, which can both strongly affect the scattered For positions A, B, C, and E, the average agreement between neutron contribution. Also, the systematic under-response of the the measurements and the simulations without the specifics of LB 6411 ambient neutron monitor could explain this the room is within 8%. The agreement for individual irradiations overestimation as can be seen in Table 3 and from the is also well within 30% for most irradiations. Taking into simulation results without the specifics of the room with account the measurement uncertainty in the range of 15% to energy correction in Table 4. 30% and the significant uncertainty on the simulations, this can For position F, it can be seen that there is a systematic already be considered as good agreement. The H*(10) values overestimation of 26% for the simulations without the specifics from the simulations with the specifics of the room in Table 4 of the room in comparison with the measurements. Adding the show a significant overestimation with respect to the specifics of the room in the simulations worsens the measurements. This overestimation is probably caused by an overestimation to about a factor of two. This is probably again TABLE 4 Overview of the average ratio of simulated over measured H*(10) per position and averaged over all positions for the Skandion measurement campaign. H*(10) Simulation/measurement A B C D E F Average Simulation without room 0.99 1.01 0.95 0.44 0.92 1.26 0.92 Simulation with room 1.54 1.13 1.50 1.23 1.58 2.04 1.50 Simulation without room with energy correction 1.36 0.86 1.37 0.21 – 0.82 1.36 Frontiers in Oncology 12 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 due to an overestimation of the scattered neutron dose neutron fluence energy spectra at positions A and F simulated in contribution because of the use of 1-m-thick concrete this work for Skandion with the simulation model with the everywhere in the gantry area in the simulation model, specifics of the room for comparison. especially since position F is completely at the back in the Very good agreement between simulations and gantry area. The overestimation is probably caused by the measurements is observed for the high-energy neutron peak. underestimation of the LB 6411 ambient neutron monitor as Only for position B is a significantly higher peak observed in the can be seen from Table 3 and from the 30% decrease of the simulations as expected because position B is 50 cm closer to the simulated H*(10) values when applying the energy correction isocenter in the simulations in comparison with the in Table 4. measurements in Trento. It can be concluded from this comparison that the The evaporation neutron peaks for positions C and D are simulation model without specifics of the room is sufficiently higher in the measurements in Trento than in the simulations. realistic. The deviations between the measured and simulated For position D, the shape is also different. This is probably neutron H*(10) values were within the estimated combined caused by simplifications in the room model or differences in uncertainty for all positions except for position D further away the rooms of Skandion and Trento. Evaporation neutrons are from the isocenter and perpendicular to the beam direction partly created in iron-rich components such as the gantry where scattered neutrons form an important dose contribution. cylinders, the gantry cone, and the counterweight (18), and Therefore, it is expected that also for the simulation of out-of- these components were not taken into account in the simplified field neutron doses within the phantom, this simulation model room model of the simulations. However, the contribution of will perform with sufficiently good accuracy. this peak to the total dose is less important closer to the phantom and inside the phantom as demonstrated by the 3.2.3 Neutron fluence energy spectrum good agreement between measured and simulated H*(10) The neutron fluence energy spectra are very important as values in Section 3.2.2. both the neutron fluence to dose equivalent conversion The thermal neutron peaks in the simulations are coefficient and the ambient neutron monitor H*(10) energy systematically higher than those of the measurements in responses depend strongly on the neutron energy. Therefore, Trento. This is probably again caused by simplifications in the verification of the simulated neutron fluence energy spectra was room model in the simulations or differences between the rooms also performed by comparison with Bonner sphere in Skandion and Trento. As discussed already in Section 3.2.2, measurements performed at a previous measurement the 1-m-thick concrete everywhere probably overestimates the campaign at a very similar proton therapy facility in Trento scattered neutron contribution and thus also the thermal (11). The setup was very similar to the measurement campaigns neutron peak. Anyhow, thermal and even epithermal neutrons at Skandion and CCB. The measurements in Trento were are expected to contribute only to the maximum, a few percent performed during an irradiation with 20 cm range, 10 cm of the total out-of-field neutron dose in proton therapy (12). modulation width, and 10 cm × 10 cm field size without a In Section 3.2.2, it was shown that the use of the simulated range shifter. The only difference was that position B in Trento neutron fluence energy spectra to correct for the imperfect was at a distance of 1.5 m from the isocenter instead of 1 m at energy response of the ambient neutron monitors can explain Skandion and CCB. some of the deviations between the simulated and the measured Figure 7 compares the neutron fluence energy spectra at H*(10) values. This gives further confidence that the simulated positions B, C, D, and E simulated in this work for Skandion neutron fluence energy spectra are sufficiently realistic. with the simulation model with the specifics of the room (thick lines) and the neutron fluence energy spectra measured with Bonner spheres at the equivalent positions in Trento (thin lines). 3.3 Variations of ambient The treatment plan parameters were the same in both data sets: dose equivalent 20 cm range, 10 cm modulation width, and 10 cm × 10 cm field size without a range shifter. The fluence energy spectra are given In Section 3.2, it was shown that the ambient neutron doses per unit of absorbed dose in the target in Gy and plotted per unit simulated with the MC model agree with the measurements lethargy. The neutron lethargy is defined as ln( EEi−1i ) = 0:26 with within the uncertainties and thus that one can rely on the results Ei and Ei−1 being the upper and lower energies of the energy bin, of the simulations. In this subsection, the simulation results are respectively. Neutron fluence energy spectra are commonly used for analysis of the variation of the neutron ambient dose plotted per unit lethargy when the energy axis is logarithmic equivalent H*(10) as a function of position inside the room and because, in this way, equal areas under the spectra represent treatment plan parameters. The simulations with the equal amount of fluence. The top plot shows all positions, while geometrical model without specifics of the room were used in the bottom plot zooms in on positions C, D, and E to have a order to make the results independent of the exact room more detailed view. The bottom plot additionally shows the geometry. The trends observed in the simulations are Frontiers in Oncology 13 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 7 Plots comparing the neutron fluence energy spectra at positions (B–E) simulated in this work for Skandion (thick lines) and the neutron fluence energy spectra measured with Bonner spheres at the equivalent positions in Trento (thin lines). The treatment plan parameters were the same in both data sets: 20 cm range, 10 cm modulation width and 10 cm × 10 cm field size without a range shifter. The fluence energy spectra are given per unit of absorbed dose in the target in Gy and plotted per unit lethargy. The top plot shows all positions, while the bottom plot zooms in on the positions (C–E) to have a more detailed view. The bottom plot additionally shows the neutron fluence energy spectra at positions A and F simulated in this work for Skandion for comparison. compared with those of the measurements performed at increasing distance from the isocenter and H*(10) is lower Skandion and, where available, also with the measurements in backward direction than in forward direction of the proton performed at CCB IFJ PAN. All the results are presented in beam. H*(10) values between 0.04 μSv and 292 μSv per Gy terms of ambient dose equivalent H*(10) per unit of absorbed target dose were obtained in the simulations. The highest dose in the center of the SOBP. doses were found for position B with an average overall treatment plan parameters of about 66 μSv per Gy. This was 3.3.1 Position expected as it is the closest position at only 1 m from the The first important factor influencing neutron H*(10) is the isocenter and located in the forward direction of the beam. position inside the room. Figure 8 shows the variation of the Lowest doses were found for positions D and E with on simulated H*(10) values (bars) and the measured H*(10) values average over all treatment plan parameters of about 1 and 3 (crosses) at Skandion as a function of the position. Average μSv per Gy, respectively. Also, this was expected as these (green), minimum (blue), and maximum (red) values are shown. positions are furthest away at 2.25 m distance in directions The positions are ordered from left to right according to perpendicular and backward with respect to the proton beam increasing distance from the isocenter and for the same direction for positions D and E, respectively. Finally, it can be distances according to increasing angle with respect to the seen that the measurements at position D are significantly proton beam direction. higher than the simulations. This is probably due to the The observed trends for simulations and measurements m i s s i n g sc a t t e r ed ne u tr o n s fr o m t h e r o o m i n th e are very similar. As expected, H*(10) decreases with simulations as discussed in Section 3.2.2. Frontiers in Oncology 14 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 8 Plot of the variation of the simulated H*(10) values (bars) and the measured H*(10) values (crosses) at Skandion as a function of the position. Average (green), minimum (blue), and maximum (red) values are shown. The positions are ordered from left to right according to increasing distance from the isocenter and for same distances according to increasing angle with respect to the proton beam direction. 3.3.2 Field size for all positions in one plot together with a linear fit through all The second factor influencing neutron H*(10) is the field data points. In this plot, the H*(10) is for each position size. Figure 9 shows the variation of the simulated H*(10) (red normalized to the H*(10) for the lowest field size for that bars connected with lines) and measured H*(10) (green and blue position. The fit has no physical meaning, but can be used as crosses) as a function of the field size for a range of 15 cm and a an approximate scaling law for modeling the dependence of the modulation width of 10 cm without a range shifter for the neutron H*(10) on the field size. different positions inside the room. The error bars on the The observed trends in simulations and measurements at measurement data points represent the k = 1 measurement Skandion and CCB are very similar. Only for position D are the uncertainties. Figure 10 shows the same simulation data but measurements significantly higher than the simulations. This is FIGURE 9 Plots of the variation of the simulated H*(10) (red bars) and measured H*(10) (green and blue crosses) as a function of the field size for a range of 15 cm and modulation width of 10 cm without a range shifter for the different positions. Frontiers in Oncology 15 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 10 Plot of the variation of the simulated H*(10) as a function of the field size for all positions for a range of 15 cm and a modulation width of 10 cm without a range shifter with a linear fit through all the data points. probably due to the missing scattered neutrons from the room in The observed trends in the simulations and measurements at the simulations as discussed in Section 3.2.2. The neutron H* Skandion are very similar. Only for position D are the (10) increases linearly with field size area for all positions. The measurements significantly higher than the simulations. This H*(10) increases with a factor of about 100 when changing from is probably due to the missing scattered neutrons from the room a field size of 3 cm × 3 cm to a field size of 30 cm × 30 cm. Inside in the simulations as discussed in Section 3.2.2. For the 10-cm the phantom, the treatment volume is closer and the neutron modulation width, there are also measurements performed at source deviates more strongly from a distant point source. CCB that are very well in line with the measurements at Therefore, inside the phantom, there might be a deviation Skandion and the simulations. It can be seen that the H*(10) from this linearity with an additional position dependence. increases relatively linearly with the range. In this case, a general The simulations also showed that the neutron energy linear scaling law does not reproduce the trends for all spectrum does not vary significantly with changes in the field modulation widths and positions properly. The increase in H* size area. (10) as a function of the range depends significantly on the position. This is probably related to the difference in distance to 3.3.3 Range the treatment volume as the main neutron source for different The next factor influencing the neutron H*(10) is the proton positions. When changing the range from 10 cm to 25 cm, the range and the corresponding proton energy. Figure 11 shows the H*(10) increases with a factor of two to eight, depending on the variation of the simulated H*(10) (red bars connected with lines) position and modulation width. and measured H*(10) (green crosses) as a function of the range The simulations also showed that the neutron fluence for a field size of 10 cm by 10 cm and a modulation width of 5 cm energy spectrum has a shift of the high-energy neutron peak without a range shifter for the different positions. The error bars towards higher energies and a decrease of the thermal on the measurement data points represent the k = 1 neutron contribution with an increase in range. This is measurement uncertainties. Figure 12 shows the same expected because an increased range means increased proton simulation data for all positions in one plot together with the energy and thus also an increase of the energy of the data for the other modulation widths and a linear fit through all high-energy neutrons and a decrease of the fraction of data points. In this plot, the H*(10) is for each position and thermalized neutrons. modulation width normalized to the H*(10) for the lowest range for that position and modulation width. The fit has no physical 3.3.4 Modulation width meaning, but can be used as an approximate scaling law for A fourth factor influencing the neutron H*(10) is the modeling the dependence of the neutron H*(10) on the modulation width. Figure 13 shows the variation of the proton range. simulated H*(10) (red bars connected with lines) and Frontiers in Oncology 16 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 11 Plots of the variation of the simulated H*(10) (red bars) and measured H*(10) (green crosses) as a function of the range for a field size of 10 cm by 10 cm and a modulation width of 5 cm without a range shifter for the different positions. measured H*(10) (green and blue crosses) as a function of the data for the other ranges and a quadratic fit through all data modulation width for a range of 20 cm and a field size of 10 cm points. In this plot, the H*(10) is for each position and range by 10 cm without a range shifter for the different positions. The normalized to the H*(10) for the lowest modulation width for error bars on the measurement data points represent the k = 1 that position and range. The fit has no physical meaning, but can measurement uncertainties. Figure 14 shows the same be used as an approximate scaling law for modeling the simulation data for all positions in one plot together with the dependence of the neutron H*(10) on the modulation width. FIGURE 12 Plot of the variation of the simulated H*(10) as a function of the range for all modulation widths and positions for a field size of 10 cm by 10 cm without a range shifter with a linear fit through all the data points. Frontiers in Oncology 17 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 FIGURE 13 Plots of the variation of the simulated H*(10) (red bars) and measured H*(10) (green and blue crosses) as a function of the modulation width for a range of 20 cm and a field size of 10 cm by 10 cm without a range shifter for the different positions. The observed trends in the simulations and measurements at the simulations as discussed in Section 3.2.2. It can be seen that Skandion and CCB are very similar. Only for position D are the the H*(10) increases quadratically with the modulation width measurements significantly higher than the simulations. This is and that the increase becomes less steep for higher modulation probably due to the missing scattered neutrons from the room in widths. A quadratic fit reproduces the trends for different ranges FIGURE 14 Plot of the variation of the simulated H*(10) as a function of the modulation width for all ranges and positions for a field size of 10 cm by 10 cm without a range shifter with a quadratic fit through all the data points. Frontiers in Oncology 18 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 and positions quite well, but there is a limited dependence on simulated without a range shifter for a range corresponding to range and position. Similar to the range, this is probably related the sum of the range with the range shifter and the solid water to the difference in distance to the treatment volume as the main equivalent thickness of the range shifter. neutron source for different positions. The H*(10) increases The simulations also showed that, similar to an increase of between 35% and 60% when increasing the modulation width the range, the use of a range shifter leads to a shift of the high- from 3 cm to 20 cm, depending on the range and position. energy neutron peak towards higher energies and a decrease of The simulations also showed that the neutron energy the thermal neutron contribution. spectrum does not change significantly when changing the Also, the dependence of the neutron H*(10) on the air gap modulation width. between the range shifter and the phantom was studied, in both the simulations and the measurements. The error bars on the 3.3.5 Range shifter and air gap measurement data points represent the k = 1 measurement The last investigated factor influencing the neutron H*(10) is uncertainties. Figure 15 shows the variation of the simulated the use of a range shifter. It was found both in the measurements H*(10) (red bars connected with lines) and measured H*(10) and the simulations that the H*(10) increases on average by a (green crosses) as a function of the air gap between the range factor of about two when introducing a range shifter for the shifter and the phantom for a range of 10 cm, a modulation same treatment. The increase was higher for lower ranges. For all width of 5 cm, and a field size of 10 cm × 10 cm for the different positions except for position D, the increase is limited to a factor positions. Both the simulations and the measurements at of 2.5, while for position D, an increase up to a factor of 6 was Skandion showed no significant effect of the variation of the observed. This is probably due to the fact that the range shifter is air gap on the neutron H*(10) for the investigated positions. positioned within line of sight from position D. The absorbed Further research is necessary to evaluate the potential influence dose measured at position A with the Sievert was also shown to of the air gap at other positions such as inside the phantom, increase up to a factor of nine when introducing a range shifter. closer to the range shifter. The absorbed dose measured at This increase is probably not due to neutrons but due to protons position A with the Sievert was shown to increase with from the range shifter as discussed in detail in (23). increasing air gap. This increase is probably not due to These observations are within expectations. When neutrons but due to protons from the range shifter as performing the same treatment, introducing a range shifter discussed in detail in (23). means increasing the proton energies as can be seen in Table A.1 in the appendix. This leads to an additional neutron creation in the range shifter. As a first approximation, one can consider 4 Conclusions and outlook that the range is increased by the solid water equivalent thickness of the range shifter. It can be seen indeed that the H*(10) values A general MC radiation transport model was set up for simulated with the range shifter are close to the H*(10) values simulation of neutron doses from scattered and secondary FIGURE 15 Plots of the variation of the simulated H*(10) (red bars) and measured H*(10) (green crosses) as a function of the air gap between the range shifter and the phantom for a range of 10 cm, a modulation width of 5 cm, and a field size of 10 cm by 10 cm for the different positions. Frontiers in Oncology 19 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 radiation at PBS proton therapy facilities. This model was μSv/Gy at 1.5 m from the isocenter perpendicular to the beam successfully verified by comparison of simulated and measured direction. The out-of-field neutron dose for a specific treatment ambient neutron doses for the irradiations of a solid water in an organ at a certain distance from the isocenter can already phantom at two different PBS proton therapy facilities using a be roughly estimated based on the dose data and scaling laws variety of treatment plans. The simulated SOBP profile properly provided in this work in combination with the inverse square reflected the range and modulation width of the plans. distance law to very roughly model the dependence on the Furthermore, the deviations between the simulated neutron H* distance from the isocenter. As a first test, this was done for (10) values and ambient neutron monitor measurements at six the recently published extensive study of out-of-field neutron different positions around the solid water phantom were well doses for a proton therapy brain treatment (24). The brain within the expected uncertainties. Finally, the simulated neutron treatment in (24) can be approximated by an irradiation with 33 fluence energy spectra were in good agreement with Bonner cm² field size, 5 cm modulation width, and 18 cm range. The sphere measurements performed during a previous irradiations in this treatment were performed under three measurement campaign with a similar setup. For the positions different angles. The out-of-field neutron doses per unit of close to the phantom where scattered neutrons from the room target dose assessed in (24) by means of measurements and do not contribute significantly to the dose, the agreement MC radiation transport simulations ranged between 10 μSv/Gy between simulated and measured neutron H*(10) was within at 50 cm from the isocenter and 1,000 μSv/Gy at 10 cm from the 30%. Therefore, it is expected that with this general MC isocenter. The most comparable dose data from this work are the radiation transport model of PBS proton therapy facilities, it dose data for position B (in beam direction at 1 m from the will also be possible to simulate out-of-field doses inside the isocenter) and position F (perpendicular to beam direction at phantom with a similarly good accuracy. 1.5 m from the isocenter) for the irradiation with 20 cm range, The MC simulations facilitated a detailed study of the 5 cm modulation width, and 100 cm² field size. Based on the variation of neutron H*(10) with position inside the room, field scaling laws obtained in this work, these dose data were divided size, range, modulation width, use of a range shifter, and air gap by a factor of three to correct for the larger field size and between the range shifter and the phantom. The neutron H*(10) multiplied with a factor of 0.84 to correct for the larger range. depends strongly on the position inside the room with a general Then, they were recalculated to the distances of 10 cm and 50 cm decrease of the neutron H*(10) with increasing distance from the by means of the inverse square distance law. This resulted in isocenter and higher neutron H*(10) in the forward direction of doses of 70 μSv/Gy (in beam direction) and 10 μSv/Gy the proton beam in comparison with the backward direction. The (perpendicular to beam direction) at 50 cm and 1700 μSv/Gy linear increase with field size and the increase of up to a factor of (in beam direction) and 300 μSv/Gy (perpendicular to beam eight with increasing range were found to be the strongest direction) at 10 cm. These doses are very well in line with the 10 influences on the neutron H*(10). The neutron H*(10) was also μSv/Gy at 50 cm and 1000 μSv/Gy at 10 cm obtained in (24). found to increase by up to about 60% with increasing modulation This example demonstrates that rough estimates of out-of-field width. The use of a range shifter on average increases the H*(10) neutron doses can already be obtained easily based on the by a factor of two. The air gap between the range shifter and the ambient neutron dose data and scaling laws provided in phantom did not have a significant influence on the neutron H* this work. (10) at the investigated positions. Further research is needed to More accurate assessment of out-of-field neutron doses evaluate potential influence of the air gap inside the phantom, requires a more sophisticated approach that will be explored closer to the range shifter. Furthermore, it was found that the in future work. It is planned to use this verified MC radiation variations of the neutron H*(10) with the treatment plan transport model for PBS proton therapy facilities to directly parameters have interdependencies and also depend on the simulate out-of-field neutron doses inside the phantom instead position inside the room. This inhibits the use of simple scaling of ambient neutron doses outside the phantom. The phantom factors to predict the H*(10) more precisely than within a factor of will be filled with 2-mm-sized boxes to allow high spatial about three close to the phantom. More precise prediction of the resolution close to the field where large dose gradients can be neutron H*(10) requires simulations at the location of interest. expected. Fluence energy spectra will be tallied in these boxes for Finally, this work demonstrates that, when reporting on out-of- neutrons, as well as for protons and photons that can also field neutron doses in proton therapy, it is important not only to contribute significantly to the out-of-field doses. These fluence normalize the out-of-field neutron doses to the target dose or energy spectra will then be convoluted with fluence to dose product of target dose with treatment volume, but also to provide equivalent conversion coefficients from literature to obtain the the treatment plan parameters. This is crucial in order to be able to out-of-field doses per unit of target dose in terms of dose compare results from different studies. equivalent. The updated simulation model will first be verified The ambient neutron doses per unit of target dose in this again for a few specific sets of treatment parameters by study varied between 3 μSv/Gy and 300 μSv/Gy at 1 m from the comparison with a previous in-phantom measurement isocenter in the beam direction and between 0.5 μSv/Gy and 50 campaign performed within EURADOS WG9 at the Trento Frontiers in Oncology 20 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 proton therapy facility. After this verification, the simulations of measurement data, manuscript review and editing. will be performed for a large series of relevant field sizes, ranges, CD: experimental setup, analysis of measurement data. MR-E: modulation widths, and range shifters. The simulations will be experimental setup, analysis of measurement data. performed both with beam direction along the long and the PO: experimental setup, analysis of measurement data. short axis of the phantom. The simulated out-of-field doses will RH: conceptualization, experimental design, manuscript review then be used to build a library and a corresponding look-up tool and editing. PO: conceptualization, experimental design, to allow assessment of the out-of-field doses at PBS proton experimental setup, manuscript review and editing. All authors therapy facilities as a function of treatment plan parameters and contributed to the article and approved the submitted version. position with respect to the isocenter and beam direction. The rectangular field treatment plans and the solid water phantom used in these simulations are of course simplified in comparison with the actual treatment of a real patient. However, actual Funding treatment plans can be simplified as one or a combination of several rectangular treatment plans. In the first instance, This project has received funding from INSPIRE from the different organs can be approximated as a series of boxes in European Union’s Horizon 2020 research and innovation the solid water phantom at a representative distance and angle programme under grant agreement No (730983) IM-R with respect to the isocenter and beam direction. Later on, the acknowledges the financial support from the Spanish Ministry simulations in the solid water phantom can be made more of Science, Innovation and Universities (RYC2018-024043-I) realistic by performing simulations in a series of representative The work of Ondrej Ploc on the paper was funded by EU anthropomorphic phantoms including the actual organs. In this Operational Program Research, Development, and Education, way, the library and tool can be extended continuously and be call 02_15_003 in project CRREAT, number CZ.02.1.01/0.0/0.0/ made more realistic over time to allow, for instance, also taking 15_003/0000481. into account differences between pediatric, adult, and pregnant patients. The tool that will be developed in this way will enable the optimization of treatment planning in terms of out-of-field Acknowledgments doses and associated detrimental effects on healthy tissue. This work was carried out within the European Radiation Dosimetry Group (EURADOS, WG9 Radiation Dosimetry Data availability statement in Radiotherapy). The original contributions presented in the study are included in the article/supplementary material. Further Conflict of interest inquiries can be directed to the corresponding author. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could Author contributions be construed as a potential conflict of interest. OV: conceptualization, experimental setup, analysis of measurement data, Monte Carlo simulation design, setup, Publisher’s note execution and analysis, manuscript writing, review and editing. LS: conceptualization, experimental design and setup, analysis All claims expressed in this article are solely those of the of measurement data, manuscript writing, review and editing. authors and do not necessarily represent those of their affiliated JL: experimental setup, analysis of measurement data, manuscript organizations, or those of the publisher, the editors and the writing, review and editing. LE: experimental setup, analysis reviewers. Any product that may be evaluated in this article, or of measurement data, manuscript review and editing. NM: claim that may be made by its manufacturer, is not guaranteed conceptualization, experimental design and setup, analysis of or endorsed by the publisher. measurement data. ML: experimental design and setup. AA: experimental design and setup, manuscript review. VM: conceptualization, experimental design and setup, analysis Supplementary material of measurement data, manuscript review and editing. FT: experimental setup, analysis of measurement data, manuscript The Supplementary Material for this article can be found online review and editing. ST: experimental setup, analysis at: https://www.frontiersin.org/articles/10.3389/fonc.2022.903537/ of measurement data. IM-R: experimental setup, analysis full#supplementary-material Frontiers in Oncology 21 frontiersin.org Van Hoey et al. 10.3389/fonc.2022.903537 References 1. Loeffler J, Durante M. Charged particle therapy–optimization, challenges and 14. Benjamin PW, Kemshall CD, Redfearn J. A high resolution spherical future directions. Nat Rev Clin Oncol (2013) 107:10, 411–424. doi: 10.1038/ proportional counter. Nucl Instr Meth (1986) 59:77–85. doi: 10.1016/0029-554X(68) nrclinonc.2013.79 90347-9 2. Aliyah F, Pinasti SG, Rahman AA. Proton therapy facilities: An overview of 15. Farah J, De Saint-Hubert M, Mojżeszek N, Chiriotti S, Gryzinski M, Ploc O, the development in recent years. IOP Conf Ser: Earth Environ Sci (2021) 927 et al. Performance tests and comparison of microdosimetric measurements with (1):012042. doi: 10.1088/1755-1315/927/1/012042 four tissue-equivalent proportional counters in scanning proton therapy. Radiat 3. Hälg RA, Schneider U. Neutron dose and its measurement in proton Measurements; (2017) 96:42–52. doi: 10.1016/j.radmeas.2016.12.005 therapy–current state of knowledge. Br J Radiol (2020) 93:20190412. doi: 16. Olsher RH, Hsu H, Beverding A, Kleck JH, Casson WH, Vasilik DG, et al. 10.1259/bjr.20190412 WENDI: an improved neutron rem meter. Health Phys (2000) 79:170–81. doi: 4. De Saint-Hubert M, Farah J, Klodowska M, Romero-Expó sito MT, Tyminska 10.1097/00004032-200008000-00010 K, Mares V, et al. The influence of nuclear models and Monte Carlo radiation 17. Werner CJ, Bull JS, Solomon CJ, Brown FB, McKinney GW, Rising ME, transport codes on stray neutron dose estimations in proton therapy. Radiat Meas et al. MCNP version 6.2 release notes, Los alamos national laboratory report LA-UR- (2022) 150:106693. doi: 10.1016/j.radmeas.2021.106693 18-20808. United States: Los Alamos National Lab (LANL) (2018). 5. Mojżeszek N, Farah J, Kłodowska M, Ploc O, Stolarczyk L, Waligó rski MPR, 18. Englbrecht FS, Trinkl S, Mares V, Rühm W, Wielunski M, Wilkens JJ, et al. et al. Measurement of stray neutron doses inside the treatment room from a proton A comprehensive Monte Carlo study of out-of-field secondary neutron spectra in a pencil beam scanning system. Phys Med (2017) 34:80–4. doi: 10.1016/ scanned-beam proton therapy gantry room. Z Med Phys (2021) 31:215–28. doi: j.ejmp.2017.01.013 10.1016/j.zemedi.2021.01.001 6. ICRP. The 2007 recommendations of the international commission on 19. Trinkl S, Mares V, Englbrecht FS, Wilkens JJ, Wielunski M, Parodi K, radiological protection. Ann (2007) 37(2-4). doi: 10.1016/j.icrp.2007.10.003 et al. Systematic out-of-field secondary neutron spectrometry and dosimetry in 7. Burgkhardt B, Fieg G, Klett A, Plewnia A, Siebert BRL. The neutron fluence pencil beam scanning proton therapy. Med Phys (2017) 44:5. doi: 10.1002/ and H*(10) response of the new LB 6411 remcounter. Radiat Prot Dosim (1997) mp.12206 70:361–4. doi: 10.1093/oxfordjournals.rpd.a031977 20. McConn RJ Jr, Gesh CJ, Pagh RT, Rucker RA, Williams RG. Compendium of 8. De Smet V. Neutron measurements in a proton therapy facility and material composition data for radiation transport modeling. Pacific Northwest comparison with Monte Carlo shielding simulations. PhD thesis. Université Libre National Laboratory report: Richland, WA, USA (2011). de Bruxelles: Brussels (2016). 21. Brown DA, Chadwick MB, Capote R, Kahler AC, Trkov A, Herman 9. Mares V, Sannikov AV, Schraube H. Response functions of the andersson-Braun MW, et al. ENDF/B-VIII.0: The 8th major release of the nuclear reaction data and extended range rem counters for neutron energies from thermal to 10 GeV, nucl. library with CIELO-project cross sections, new standards and thermal Instrum Meth (2002) 476(1-2):341–6. doi: 10.1016/S0168-9002(01)01459-0 scattering data. Nucl Data Sheets (2018) 148:1–142. doi: 10.1016/ j.nds.2018.02.001 10. Trompier F, Delacroix S, Vabre I, Joussard F, Proust J. Secondary exposure for 73 and 200 MeV proton therapy. Radiat Prot Dosimetry; (2007) 125(1-4):349– 22. De Smet V, De Saint-Hubert M, Dinar N, Manessi GP, Aza E, Cassell C, 54. doi: 10.1093/rpd/ncm154 et al. Secondary neutrons inside a proton therapy facility: MCNPX simulations compared to measurements performed with a Bonner sphere spectrometer and 11. Farah J, Mares V, Romero-Expó sito M, Trinkl S, Domingo C, Dufek V, et al. neutron H*(10) monitors. Radiat Measurements (2017) 99:25–40. doi: 10.1016/ Measurement of stray radiation within a scanning proton therapy facility: j.radmeas.2017.03.005 EURADOS WG9 intercomparison exercise of active dosimetry systems. Med 23. Eliasson L, Lillhök J, Bäck T, Billnert-Maró ti R, Dasu A, Liszka M. Range- Phys (2015) 42(5):2572–2584. doi: 10.1118/1.4916667 shifter effects on the stray field in proton therapy measured with the variance- 12. Mares V, Romero-Expó sito M, Farah J, Trinkl1 S, Domingo C, Dommert M, covariance method. Submitted to Front Oncol same special issue as this work (2022) et al. A comprehensive spectrometry study of a stray neutron radiation field in scanning 12:882230. doi: 10.3389/fonc.2022.882230 proton therapy. Phys Med Biol (2016) 61:4127–40. doi: 10.1088/0031-9155/61/11/4127 24. De Saint-Hubert M, Verbeek N, Bäumer C, Esser J, Wulff J, Nabha R, et al. 13. Kyllönen J-E, Lindborg L, Samuelson G. The response of the sievert Validation of a Monte Carlo framework for out-of-Field dose calculations in instrument in neutron beams up to 180 MeV. Radiat Prot Dosim (2001) 94:227– proton therapy. Front Oncol Volume (2022) 12:882489. doi: 10.3389/ 32. doi: 10.1093/oxfordjournals.rpd.a006494 fonc.2022.882489 Frontiers in Oncology 22 frontiersin.org