Workshop on the Monte Carlo radiotherapy system PRIMO

Book of abstracts

With the support of Event endorsed by

September 20—22, 2017

2

3

Workshop on the Monte Carlo radiotherapy system PRIMO

Book of abstracts

4

5

Workshop on the Monte Carlo radiotherapy system PRIMO

Venue: Universitätsklinikum Essen Operatives Zentrum II, Auditorium Hufelandstraße 55 45147 Essen Germany Dates: September 20—22, 2017 Event supported by: Deutsche Forschungsgemeinschaft (DFG) Universitätsklinikum Essen Medizinische Fakultät der Universität Duisburg-Essen Event endorsed by: European Society for Radiotherapy & Oncology (ESTRO) Scientific committee President: Lorenzo Brualla (Universität Duisburg-Essen, Germany) Pedro Andreo (Karolinska Institutet, Sweden) Luca Cozzi (Istituto Clinico Humanitas, Italy) Michael Fix (Inselspital, Switzerland) Nuria Jornet (Hospital de la Santa Creu i Sant Pau, Spain) Antonio Lallena (Universidad de Granada, Spain) Nick Reynaert (Centre Oscar Lambret, France) Miguel Rodriguez (Centro Médico Paitilla, Panama) Wolfgang Sauerwein (Universität Duisburg-Essen, Germany) Josep Sempau (Universitat Politècnica de Catalunya, Spain) Virginia Tsapaki (Konstantopoulio General Hospital, Greece) Andrea Wittig (Universitätsklinikum Jena, Germany) Organizing committee President: Wolfgang Sauerwein (Universität Duisburg-Essen, Germany) Lorenzo Brualla (Universität Duisburg-Essen, Germany) Luca Cozzi (Istituto Clinico Humanitas, Italy) Gundula Franz (Universitätsklinikum Essen, Germany) Miguel Rodriguez (Centro Médico Paitilla, Panama) Josep Sempau (Universitat Politècnica de Catalunya, Spain) Erdin Tokmak (Universität Duisburg-Essen, Germany)

6

7

Contents

Program page 9 Do we need the Monte Carlo method in the clinical environment? The point of view of a clinician W. Sauerwein page 13 Do we need the Monte Carlo method in the clinical environment? The point of view of a medical physicist L. Cozzi page 14 Review of Monte Carlo codes applied to radiation therapy L. Brualla, M. Rodriguez, A.M. Lallena page 15 Moving PRIMO to the cloud H. Miras, R. Jimenez page 16 Monte Carlo dose calculation of VMAT clinical treatment plans for Novalis TrueBeam STx linac using PRIMO A. Sottiaux, V. Baltieri, C. Leclercq, A. Monseux, M. Tomsej page 18 Monte Carlo simulation of translational total body irradiation with PRIMO and source 20 of DOSXYZnrc K.F. Löwenich, T. Streller page 20 Evaluation of clinical plans using PRIMO Dose Planning Method (DPM) for brain and lung metastases G. Reggiori, A. Stravato, F. Zucconi, L. Paganini, P. Mancosu, V. Palumbo, S. Tomatis page 22 Evaluation of the target dose inhomogeneity in breast cancer treatment due to tissutal differences A. Fogliata, A. Stravato, G. Reggiori, M. Scorsetti, L. Cozzi page 23 Radiation dose distribution in functional heart regions—comparison of Acuros XB and PRIMO H. Karle, S. Berthes, M. Stockinger, D. Wollschlaeger, H. Schmidberger, U. Wolf page 24 Accuracy of PRIMO software validated against a reference dosimetry dataset for Varian clinical linear accelerators J.F. Calvo-Ortega, D. Sanchez-Artunedo, M. Hermida-Lopez page 27

8

A Monte Carlo study of Varian Unique linear accelerator using PRIMO software S. Irmak page 30 Monte Carlo simulation of 3D-scanner in PRIMO: comparison with experimental data R. Tortosa, C. Senra, S. Diez, R.M. Cibrian page 31 Monte Carlo simulations in radiotherapy dosimetry P. Andreo page 33 PENELOPE transport parameters for linac simulation J. Sempau, M. Rodriguez, L. Brualla page 35 Monte Carlo simulations of intensity modulated radiotherapy using PRIMO A. Esposito, J. Oliveira, S. Silva, J. Lencart, J. Santos page 37 The Dose Planning Method code J. Sempau page 40 Improvements on DPM for PRIMO M. Rodriguez page 42 What’s new in PRIMO M. Rodriguez, J. Sempau, L. Brualla page 44 Study of build-up dose region for 6 MV photon beam using PRIMO simulation software M. Mahur, P.S. Negi, M. Semal. M. Singh page 45 Determination of initial parameters of Varian 2100 CD linac electron beams using PRIMO R. Maskani, M.J. Tahmasebibirgani, M. Hoseini-Ghahfarokhi, J. Fatahiasl page 46 Dosimetric evaluation of AAA and Acuros XB algorithms with the PRIMO software in inhomogeneous media M.L. Martin-Albina, F. Pizarro Trigo, J. Jimenez Albericio, L.M.R. Nunez Martinez, J. Morillas Ruiz, J. Sanchez Jimenez page 49 A Monte Carlo study for a breast conformal radiotherapy plan using PRIMO software S. Irmak page 52

9

Program

September 20, 2017: 09:00—09:15 Welcome Lorenzo Brualla (Universitätsklinikum Essen, Germany) Session 1: Do we need the Monte Carlo method in the clinical environment? Chair: Pedro Andreo (Karolinska Institutet, Sweden) 09:15—09:45 The point of view of a clinician Wolfgang Sauerwein (Universitätsklinikum Essen, Germany) 09:45—10:15 The point of view of a medical physicist Luca Cozzi (Istituto Clinico Humanitas, Italy) 10:15—10:45 Review of Monte Carlo codes applied to radiation therapy Lorenzo Brualla (Universitätsklinikum Essen, Germany) 10:45—11:15 Coffee break Session 2: Algorithms I Chair: Lorenzo Brualla (Universitätsklinikum Essen, Germany) 11:15—11:45 Moving PRIMO to the cloud Héctor Miras (University Hospital Virgen Macarena, Spain) 11:45—12:15 Monte Carlo dose calculation of VMAT clinical treatment plans for Novalis TrueBeam STx linac using PRIMO Alain Sottiaux (CHU Charleroi, Belgium) 12:15—13:30 Lunch Session 3: Clinical Chair: Wolfgang Sauerwein (Universitätsklinikum Essen, Germany) 13:30—14:00 Monte Carlo simulation of translational total body irradiation with PRIMO and source 20 of DOSXYZnrc Karl Löwenich (University Hospital Zurich, Switzerland)

10

14:00—14:30 Evaluation of clinical plans using PRIMO Dose Planning Method (DPM) for brain and lung metastases Giacomo Reggiori (Istituto Clinico Humanitas, Italy) 14:30—15:00 Evaluation of the target dose inhomogeneity in breast cancer treatment due to tissutal differences Antonella Fogliata (Istituto Clinico Humanitas, Italy) 15:00—15:30 Dosimetric evaluation of AAA and Acuros XB algorithms with the PRIMO software in inhomogeneous media Javier Jiménez (Hospital Universitario Miguel Servet, Spain) 15:30—16:00 Coffee break Session 4: Dosimetry I Chair: Luca Cozzi (Istituto Clinico Humanitas, Italy) 16:00—16:30 Accuracy of PRIMO software validated against a reference dosimetry dataset for Varian clinical linear accelerators Marcelino Hermida (Hospital Universitari Vall d’Hebron, Spain) 16:30—17:00 A Monte Carlo study of Varian Unique linear accelerator using PRIMO software Sinan Irmak (Acibadem Kayseri Hospital, Turkey) 17:00—17:30 Monte Carlo simulation of 3D-scanner in PRIMO: comparison with experimental data Ricardo Tortosa (Hospital IMED Elche, Spain) September 21, 2017: Session 5: Keynote lecture Chair: Josep Sempau (Universitat Politècnica de Catalunya, Spain) 09:00—09:45 Monte Carlo simulations in radiotherapy dosimetry Pedro Andreo (Karolinska Institutet, Sweden) Session 6: Algorithms II Chair: Lorenzo Brualla (Universitätsklinikum Essen, Germany) 09:45—10:15 PENELOPE transport parameters for linac simulation Josep Sempau (Universitat Politècnica de Catalunya, Spain)

11

10:15—10:45 Monte Carlo simulations of IMRT using PRIMO Alessandro Esposito (Oncology Institute of Porto, Portugal) 10:45—11:15 Coffee break Session 7: New version of PRIMO Chair: Antonella Fogliata (Istituto Clinico Humanitas, Italy) 11:15—11:45: The Dose Planning Method code Josep Sempau (Universitat Politècnica de Catalunya, Spain) 11:45—12:15 Improvements on DPM for PRIMO Miguel Rodriguez (Centro Médico Paitilla, Panama) 12:15—12:45 What’s new in PRIMO Miguel Rodriguez (Centro Médico Paitilla, Panama) 12:45—14:00 Lunch Session 8: Dosimetry II Chair: Miguel Rodriguez (Centro Médico Paitilla, Panama) 14:00—14:30 Study of build-up dose region for 6 MV photon beam using PRIMO simulation software Manta Mahur (Delhi State Cancer Institute, India) 14:30—15:00 Determination of initial parameters of Varian 2100 CD linac electron beams using PRIMO Reza Maskani (Shahroud University of Medical Sciences, Iran) 15:00—15:30 Radiation dose distribution in functional heart regions— comparison of Acuros XB and PRIMO Heiko Karle (University Medical Center Mainz, Germany) 15:30—16:00 Coffee break Session 9: Ask the authors Chair: Pedro Andreo (Karolinska Institutet, Sweden) 16:00—17:30 Participants are invited to ask questions or comment difficulties on PRIMO to its authors. The participants can use slides to facilitate the discussion. Lorenzo Brualla, Miguel Rodriguez, Josep Sempau 19:00—21:00 Social dinner

12

September 22, 2017: Session 10: Teaching the new PRIMO version Instructors: Lorenzo Brualla, Miguel Rodriguez, Josep Sempau 09:00—10:30 Hands-on session for learning the new features of PRIMO 10:30—11:00 Coffee break 11:00—12:30 Hands-on session for learning the new features of PRIMO 12:30—12:45 Closing remarks Lorenzo Brualla (Universitätsklinikum Essen, Germany)

13

Do we need the Monte Carlo method in the clinical environment? The point of view of a clinician

Wolfgang Sauerwein

NCTeam. Strahlenklinik. Universitätsklinikum Essen. Hufelandstr 55. 45147 Essen. Germany. Presenting author: wolfgang.sauerwein@uni-due.de Absorbed dose is a fundamental quantity used in all therapeutic applications of ionizing radiation. Measurement and reporting of the absorbed dose is crucial to the understanding of any radiation effects. Absorbed doses can never be measured in body tissues directly, as it is unknown, how much energy is used for warming and how much for chemical reactions in cells. The dose measurement for photon and electron therapy, which both act by energy transfer via electrons, is therefore, based on conversion of the measured value of an ionization chamber, calibrated in terms of water absorbed dose, in the absorbed dose in body tissues. The necessary correction factors, including their most important energy, depending on whether they are, are well known, as long as the center of a relatively large and homogenous volume is observed. Current treatment planning systems (TPS) are mostly based on such measurements and therefore unable to accurately display the real absorbed dose in a given point or structure, if small field sizes are used, if volumes out of axes are analyzed and if high density gradients to the neighboring voxels are present. Modern radiotherapy techniques are searching for steep gradients between target areas and surrounding tissues, very small field sizes are often used especially in IMRT but inherent uncertainties of the available TPS are not really taken into consideration. Further development of external beam radiotherapy urgently needs TPS resulting a more accurate description of the absorbed dose. Monte Carlo methods seem to be a powerful tool to achieve this goal.

14

Do we need the Monte Carlo method in the clinical environment? The point of view of a medical physicist

Luca Cozzi

Humanitas Research Hospital and Cancer Center, Milan-Rozzano, Italy Presenting author: luca.cozzi@cancercenter.humanitas.it The house believes that in the process of accurate preparation of a radiotherapy course, several challenges exist but some of those can be efficiently managed with modern tools. In the lecture, the need and the relevance of the use of advanced, accurate and reliable dose calculation algorithms will be reviewed from a clinical-physicist perspective. The main pitfalls of the “old” fashioned algorithms will be analysed in the perspective of their mitigation or possible full solution with Monte Carlo or equivalent approaches. The clinical cases of breast and lung irradiation will be used as practical examples demonstrating the thesis. Some open questions and hot topics will also be mentioned to stimulate the discussion and to force us to appraise the consequences of an easy “status-quo” life compared to an adventurous navigation through stochastic seas.

15

Review of Monte Carlo codes applied to radiation therapy

Lorenzo Brualla*,1, Miguel Rodriguez2, Antonio M. Lallena3 (1) NCTeam. Strahlenklinik. Universitätsklinikum Essen. Hufelandstr 55. 45147 Essen. Germany. (2) Centro Médico Paitilla. Ciudad de Panamá. Panama (3) Departamento de Física Atómica, Molecular y Nuclear. Universidad de Granada. 18071 Granada. Spain. Presenting author: lorenzo.brualla@uni-due.de General-purpose radiation transport Monte Carlo codes have been used for the estimation of the absorbed dose distribution in external beam radiotherapy patients since several decades. Results obtained with these codes are usually more accurate than those provided by treatment planning systems based on non-stochastic methods. Absorbed dose computations based on general-purpose Monte Carlo codes have been traditionally used only for research owing to the difficulties associated to setting up a simulation and the long computation time required. In an effort to extend the application of Monte Carlo codes to the routine clinical practice researchers and private companies have developed treatment planning and dose verification systems that are partly or fully based on fast Monte Carlo algorithms. We present a comprehensive list of the currently existing Monte Carlo systems that can be used to calculate a treatment plan or to verify it. Particular attention is given to those systems that are distributed, either freely or commercially, and that do not require programming tasks from the end-user. These systems are compared in terms of features and simulation time required to compute a set of benchmarks.

16

Moving PRIMO to the Cloud

Héctor Miras*,1,2, Rubén Jimenez3

(1) University Hospital Virgen Macarena, Medical Physics Department, Seville, Spain. (2) IBIS, Institute of Biomedicine of Seville, Seville, Spain. (3) Icinetic TIC SL, R&D division, Seville, Spain. (*) Presenting author: hector.miras@gmail.com Purpose/Objective PRIMO is a software solution that allows to perform easy Monte Carlo (MC) simulations of most Varian and Elekta LINACs. Despite the possibility of using different variance reduction techniques, the time required to simulate a complete radiotherapy (RT) treatment in PRIMO in an ordinary computer is still too long to use it in clinical routine. In this work we studied the possibility of moving PRIMO to a Cloud Computing platform specifically designed to perform MC simulations of RT treatments in a fast, easy and economical way. Material/Methods The Cloud Computing platform used is CloudMC1, which has been developed by the authors of this work. CloudMC is a cloud application that runs on the commercial cloud Microsoft Azure. It has been designed to reduce MC computing times by means of parallelization in a build-on-demand virtual machines cluster. The last features implemented in CloudMC include the possibility of running MC calculations of complete RT treatments. A MC model has been created in CloudMC combining PenEasyLinac and PenEasy programs. PenEasyLinac transcribes the geometric information of the treatment beams (MLC, Jaws, etc.) from the RTPLAN to the PenEasy geometry input files. Then, a first run of PenEasy generates a phase space file that will be used as source in a second run on the voxelized geometry built from the CT images of the patient. This sequence of actions is performed automatically, users just need to upload the RTPLAN and the CT image set in DICOM format to obtain a MC calculation of a RT treatment. The results are output in RTDOSE DICOM format and can be compared with those obtained with a commercial planning system. Results Preliminary results of CloudMC simualtions of step & shoot IMRT plans, designed for several Varian and Elekta LINAC models, show that it is possible to reduce computing times to few minutes when using enough parallel computing nodes in the Cloud. Firsts tests performed using 200 computing instances take 15-30 min to simulate an IMRT treatment, in comparison to >50 h in a single core ordinary PC.

17

Conclusions PRIMO project can be implemented easily in CloudMC. The combination of these two projects makes it possible to calculate RT treatments from a wide variety of LINACs with PENELOPE MC code in short times. References 1H Miras, R Jiménez, C Miras and C Gomà. CloudMC: a cloud computing application for Monte Carlo simulation. Phys.Med.Biol. 58 (2013) N125-N133.

18

Monte Carlo dose calculation of VMAT clinical treatment plans for Novalis TrueBeam STx linac using PRIMO

Alain Sottiaux*,1,2, Valérie Baltieri1,2, Cédric Leclercq1,2,

Anne Monseux1,2, Milàn Tomsej1,2 (1) CHU de Charleroi, Hospital A. Vésale (2) CERAH (Centre de Radiothérapie du Hainaut) (*) Presenting author: alain.sottiaux@chu-charleroi.be I. Introduction and purpose With increasing complexity irradiating techniques like VMAT, QA becomes more important and challenging. Patient dedicated QA both focus on dose calculation and dose delivery verification. PRIMO was tested for clinical VMAT plans on patient CT, delivered with Varian TrueBeam STx linac. II. Material and methods II.A. PRIMO PRIMO (V0.1.5.1300) is a Monte Carlo dose calculation software that simulates radiotherapy Linacs. It estimates absorbed dose distributions in water/slab phantoms and CT. It combines a graphical user interface, the geometry of some Linacs and their MLC, and a computation engine based on the Monte Carlo code PENELOPE. PRIMO (V0.3.1.1363, Beta release) includes VMAT capabilities with DPM Monte Carlo code. As no publication of the exact geometry of TrueBeam Linac’s head can be found in the literature, we used generic phase space files made available by Varian. PRIMO calculations were performed, based on these phase spaces in order to check if they fit measurement made on our Linac. II.B. In-house Python scripts PRIMO 0.1.5 doesn’t implement VMAT capabilities and thus some workarounds are required. Python (v3.5) scripts have been therefore developed to: • Convert DICOM RTPlan in PRIMO project definition format, • Merge phase spaces before dose deposit in CT (speed up calculation, take into account variable dose rate), • Convert calculated dose from PRIMO to DICOM RTDose. PRIMO dose (in eV/g/initial history) is converted to Gy with a conversion factor calculated from a reference field. PRIMO 0.3.1 only requires DICOM RTDose conversion II.C. Dose calculation evaluation Converted dose from PRIMO (both from 0.1.5/PENELOPE and 0.3.1/DPM) can be imported in Eclipse. Doses from PRIMO and Eclipse (AcurosXB 13.7, Dose to medium) are compared, both in terms of DVH and isodoses. npgamma (0.7.0) is a Python package for 3D gamma index evaluation from DICOM RTDose files. This allows a quantitative metric for dose comparison.

19

II.D. Results and discussion There is a good agreement between dose calculations from PRIMO and from AcurosXB, although main differences arise in air cavities. PRIMO 0.1.5, even with external workaround to provide VMAT support, still has some limitations: all CT volume is taken into account, limited number of slices, Linac couch inclusion larger than CT images impossible. PRIMO 0.3.1 allows straightforward and faster VMAT plans calculation. III. Conclusions In our department, PRIMO (V0.1.5.1300) can therefore be used for dose calculation of VMAT clinical plans, on CT, with TrueBeam STx. This requires some workaround, and still faces some limitations. Next PRIMO version (0.3.1) integrates VMAT, and offers a perspective to allow systematic independent calculation of all clinical VMAT plans. Acknowledgements and potential conflicts of interest Acknowledgements go to PRIMO team (L.Brualla, J.Sempau, M.Rodriguez) for their support, encouragements and collaboration References https://www.primoproject.net M. Rodriguez, J. Sempau and L. Brualla, PRIMO: A graphical environment for the Monte Carlo simulation of Varian and Elekta linacs, Strahlenther. Onkol. 189 (2013) 881-886.

20

Monte Carlo simulation of translational total body irradiation with PRIMO and source 20 of DOSXYZnrc

Karl F. Loewenich*, Tino Streller

Department of Radiation Oncology, University Hospital Zurich (*) Presenting author: KarlFerdinand.Loewenich@usz.ch Purpose Total body irradiation (TBI) is part of patient conditioning prior to hematopoietic stem cell transplantation. In our Institution TBI is administered using the translational couch technique where during treatment the patient, while lying on a movable couch, is transported through a static beam (SSD = 180 cm, field size 35x38 cm at isocenter)) with constant velocity. As standard treatment planning systems do not generally allow for calculations of dose distributions in translational treatment modalities, in this work the Monte Carlo (MC) codes PRIMO [1] and DOSXYZnrc [2] were employed to achieve this. Materials and Methods For MC dose calculations in LINAC based radiation therapy, precise data of geometry and materials are normally needed for the tedious work of constructing an accelerator model. In this work, we could avoid the subtleties involved in that task by taking advantage of the PRIMO code in which models of commonly used accelerators are readily incorporated. To simulate the TrueBeam accelerator, IAEA Phasespace files (PHSPF’s 27 cm from source) supplied by VARIAN were read into PRIMO which then calculated a PHSPF (70 cm from source) for the jaws setting employed in the TBI. Using this PHSPF, depth dose curves in water were calculated for verification purpose within PRIMO as well as in DOSXYZnrc and compared to experimental data. To take the presence lung shielding of the patient during TBI into account in the MC calculation, MV images, taken prior to treatment for verification of the position of the shields, were registered with patient CT-data. The position and shape of the shielding material was then detected in the MV image with image processing methods and introduced into the original CT-data using in-house software created for the purpose. The translational TBI dose distribution was then calculated using source 20 of DOSXYZnrc together with the PRIMO generated PHSPF and the CT images of the patient in prone and supine position. The total dose distribution in the patient was then obtained using deformable registration of the distributions in prone and supine position.

21

Results PRIMO and DOSXYZnrc MC calculations and measurements in water phantom showed reasonable agreement. Source 20 of DOSXYZnrc could be employed for calculation of dose distribution in translational couch TBI. Lead absorbers used on the patient body surface could be successfully introduced into the MC calculations allowing to assess the reliability of lung dose reduction by the absorbers. Conclusion Dose distributions for translational TBI with extended SSD fields may be predicted by MC calculations which can be used for independent monitor unit calculations as well as to estimate the dose to organs of risk, Especially the important inhomogeneous lung dose distribution may be evaluated, The calculations are feasible even with a common laptop (8 cores, 2.4GHz) in a reasonable time (10h) with sufficient accuracy. References [1] Rodriguez M, Sempau J, Brualla L. PRIMO: A graphical environment for the Monte Carlo simulation of Varian and Elekta linacs. Strahlenther Onkol DOI 10.1007/s00066-013-0415-1. [2] DOSXYZnrc User’s Manual B. Walters, I. Kawrakow and D.W.O. Rogers, Ionizing Radiation Standards National Research Council of Canada, Ottawa

22

Evaluation of clinical plans using PRIMO Dose Planning Method (DPM) for brain and lung metastases

Giacomo Reggiori*, Antonella Stravato, Fabio Zucconi, Lucia Paganini,

Pietro Mancosu, Valentina Palumbo, Stefano Tomatis Physics unit of the Radiotherapy Department. Humanitas Clinical and Research Hospital. Milan-Rozzano, Italy (*) Presenting author: giacomoreggiori@gmail.com Objective The objective of this work was to evaluate the performances of clinical algorithm (AAA and Acuros) of Varian Eclipse treatment planning system (TPS) using PRIMO Monte Carlo calculation as a benchmark. Material and Methods A set of 10 patient was selected for this study: 5 patients with brain metastases and 5 with lung metastases. For all patients VMAT plans were optimized in Eclipse and calculated both wit AAA and Acuros algorithms. The generated DICOM files (plan, structure and images) were imported in the PRIMO environment where the 10 MV FFF beam of the Varian TrueBeam was simulated using the Varian TB phsp files. The Dose Planning Method (DPM) simulation engine was then used to calculate the dose distribution in the patients CTs. The first group of plans was used to evaluate the impact of the TPS modeling parameters, in particular the backscattering factor (BSF) and the beam spot size (BSS). The second group focused on the behavior of the algorithm in highly inhomogeneous tissues. The different plans were compared in terms of DVHs (dose volume histograms) comparing various dose-volume endpoints. Results Preliminary results show an acceptable agreement between Monte Carlo simulations and Eclipse plans. Acuros confirms an improved ability in dealing with inhomogeneities compared to AAA. Conclusion PRIMO is an interesting tool for the clinical environment useful to verify and support the commercially available TPS. It is an affordable Monte Carlo environment even in hospital departments where low IT resources are available.

23

Evaluation of the target dose inhomogeneity in breast cancer treatment due to tissutal differences

Antonella Fogliata*, Antonella Stravato, Giacomo Reggiori,

Marta Scorsetti, Luca Cozzi

Humanitas Research Hospital and Cancer Center, Milan-Rozzano, Italy Humanitas University, Dept of Biomedical Sciences, Milan-Rozzano, Italy (*) Presenting author: antonella.fogliata@humanitas.it Objective The mammary gland consists of many small lobules composed of connective tissue compartments (in average about 40% of the entire breast), separated by adipose tissue (about 60% of the breast volume). This difference is reflected with different material assignments for Monte Carlo calculations, being the lobular fraction can be associated to muscle, while the surrounding fat can be associated to adipose tissue. Aim of the present study is the dosimetric behaviour derived from this complex composition, in terms of target (breast) dose homogeneity. Method Five breast patients were planned with different techniques: 3D-CRT with field in field, fixed beams intensity modulation IMRT, volumetric modulated arc therapy with 6 MV beams from a Varian TrueBeam linac. Plans were optimized with the Varian Eclipse treatment planning system (version 13.6) and exported together with CT dataset and Dicom structures. Those were then imported in PRIMO for Monte Carlo calculations (using FakeBeam unit and the Varian phase space files). The same plans were also calculated with Acuros XB (Linear Boltzmann transport equation solver), using the same chemical composition assignment for HU ranges as in PRIMO. Dose distribution in the lobular and in the fat regions of the planning target volume PTV (delineated as PTV_lob and PTV_fat) was estimated and compared. Comparison between PRIMO Monte Carlo and Acuros were assessed. Results The analysis of the two breast stuctures with densities compatible with muscle and adipose tissues showed a systematic difference in the dose calculation, for the preliminary results, of about 2%, being higer in the lobular breast density regions. Dose differences were extimated also with Acuros. This unavoidable dose inhomogeneity is quantified for the different planning techniques. Conclusion Monte Carlo accuracy and dose to medium estimation allows to better understand the dose deposited in the various part of a target, depending on their specific tissutal composition. The difference is found to be significant. Due to the small absolute difference is hard to transfer this improved information into possible clinical outcome.

24

Radiation dose distribution in functional heart regions—comparison of Acuros XB and PRIMO

Heiko Karle*,1, Sahra Berthes1,2, Marcus Stockinger1, Daniel Wollschläger3,

Heinz Schmidberger1, Ulrich Wolf4

(1) Department of Radiation Oncology, University Medical Center Mainz (2) Department of Physics, Heinrich-Heine-University Duesseldorf (3) Institute for Medical Biostatistics, Epidemiology and Informatics; University Medical Center Mainz (4) Department of Radiation Oncology, University Leipzig (*) Presenting author: karle@uni-mainz.de Purpose/Objective To quantify the effect of different dose calculation algorithms and their impact on out-of-field dose estimations in functional heart regions from tangential breast irradiation. Material and methods Breast cancer radiotherapy was shown to increase the risk of death from heart disease in retrospective epidemiologic cohort studies. Typically, these studies used estimations of mean heart dose to quantify a dose effect relationship1. In newer trials, commercial treatment planning systems are used for dose estimations purposes2. For a more detailed analysis of the dose to the heart in patients undergoing tangential breast irradiation we defined seven functional regions (complete heart, aortic valve, pulmonary valve, myocardium, left ant. myocardium, right ant. myocardium and the AV node) according to the PASSOS heart atlas [Fig 1]3. Dose calculations were done for five patients with left-sided breast cancer using Varian´s Acuros XB (Eclipse V. 13.1) and PRIMO. The CT-scans had a spatial resolution between 3mm and 10mm. Treatment plans were calculated for all patients with 6MV tangential fields using a field in field technique. PTV dose prescription was normalized to D50%=50Gy (dose to medium). Results The mean heart dose was 3,0±1,7Gy for the cases calculated with Acuros XB and 3,2±1,7Gy for the calculations made with PRIMO. The values for all substructures of the heart showed the same tendency with slightly higher values for structures calculated with PRIMO compared with corresponding calculations from Varian´s Acuros XB (aortic valve 1,0±0,28Gy vs. 1,2±0,27Gy, pulmonary valve 2,2±0,72Gy vs. 2,5±0,88Gy, myocardium 4,4±2,74Gy vs. 4,5±2,68Gy, left ant. myocardium 12,7±8,58Gy vs. 12,8±8,13Gy, right ant. myocardium 1,1±0,34Gy vs. 1,3±0,36Gy and the AV node 1,1±0,22Gy vs. 1,4±0,24Gy). The biggest deviations between the calculation algorithms could be found in the low dose region receiving up to 4 Gy. The minimum doses D99% for the PTVs as calculated by PRIMO were a little lower than corresponding doses from Acuros XB (44,38±0,94Gy vs. 45,11±0,85Gy) while the PRIMO maximum doses D1ccm were a little higher than Acuros XB doses (56,33±0,63Gy vs. 55,03±1,49Gy).

25

Fig. 1.: CT slice with contours of the complete heart (orange) and functional sub-structures as defined by the PASSOS heart atlas: aortic valve (light pink), pulmonary valve (blue), right anterior heart wall (light blue), left anterior heart wall (red), and complete myocardium (purple region) - the AV node is not shown in the picture3.

Fig. 2: Typical dose volume-histograms (DVHs) for a left-sided tangential breast treatment including the PTV and the substructures of the heart calculated with Acuros XB (solid line) and PRIMO (dashed line).

26

Conclusion There are slight differences between the Acuros XB and the PRIMO based calculations when comparing the DVHs. Differences in the high dose region could be seen in heterogeneous regions with large density differences between the PTV and the lung/air. For all PASSOS-structures the mean dose was a little higher when using PRIMO – especially in the region around 4 Gy and below. Commercial treatment planning systems are not designed for calculating doses in the low-dose out of field area, and the observed differences could be explained, by their lack of accuracy in this region. It is necessary to further quantify the limits of accuracy (e.g., the impact of inter- and intra-fractional motion) for a more detailed dose analysis in functional heart regions for radiotherapy patients. Conflict of interest statement All authors declare no conflict of interests. References 1 S.C. Darby, M. Ewertz, P. McGale, A.M. Bennet, U. Blom-Goldman, D. Bronnum, C. Correa, D. Cutter, G. Gagliardi, B. Gigante, M.B. Jensen, A. Nisbet, R. Peto, K. Rahimi, C. Taylor, P. Hall, "Risk of ischemic heart disease in women after radiotherapy for breast cancer," The New England journal of medicine 368, 987-998 (2013). 2 D. Wollschlager, H. Karle, M. Stockinger, D. Bartkowiak, S. Buhrdel, H. Merzenich, T. Wiegel, M. Blettner, H. Schmidberger, "Radiation dose distribution in functional heart regions from tangential breast cancer radiotherapy," Radiotherapy and oncology : journal of the European Society for Therapeutic Radiology and Oncology 119, 65-70 (2016). 3 D. Wollschlager, H. Karle, M. Stockinger, D. Bartkowiak, S. Buhrdel, H. Merzenich, T. Wiegel, H. Schmidberger, M. Blettner, "Predicting Heart Dose in Breast Cancer Patients Who Received 3D Conformal Radiation Therapy," Health physics 112, 1-10 (2017).

27

Accuracy of PRIMO software validated against a reference dosimetry dataset for Varian clinical linear accelerators

Juan Francisco Calvo-Ortega1, David Sánchez-Artuñedo2,

Marcelino Hermida-López*,2

(1) Servicio de Radioterapia. Hospital Quirón. Barcelona. Spain. (2) Servei de Física i Protecció Radiològica. Hospital Universitari Vall d’Hebron. Barcelona. Spain. (*) Presenting author: mhermida@vhebron.net Objective The parameters that characterize the initial electron beam in the simulation of a clinical linac, mainly the mean energy and the focal spot size, largely influence on the simulated dose distributions. The users of a Monte Carlo code for simulation transport must match the simulated dose distributions to measurements and this is achieved usually in a lengthy iterative process, by tuning the simulation parameters and comparing the simulated with the measured dose distributions. To reduce the time needed for this tuning process, PRIMO suggests default values of the initial beam parameters for each nominal energy of the available linac models. This work investigates the suitability of PRIMO default beam parameters by comparing dosimetric data obtained from PRIMO simulations with a published dataset based on measurements on a large series of linacs of the same model. We focused on dosimetric data of 6 MV photon beams from Varian Clinac 2100 linacs with a Millennium 120 MLC. Material and methods We used PRIMO v. 0.1.3.1343 to obtain a phase-space file (PSF) with 450 million histories of the s1 stage of a Clinac 2100 with a 6 MV photon beam (E=5.4 MeV by default). We used this PSF to estimate dosimetric parameters in a water phantom for some setups of interest (static fields). Point measurement distributions provided by the Imaging and Radiation Oncology Core-Houston (IROC-H) Quality Assurance Center were used as benchmark data1.2. The accuracy of PRIMO simulation results was assessed for several dosimetric parameters: percentage depth doses, jaw output factors, MLC small-field output factors, and off-axis ratios. Results All evaluated dosimetric parameters obtained from the simulations with PRIMO (see Table) agree within 2% with the data measured by IROC-H, except the SBRT-style output factors, which agree within 3%. Although a fine-tuning is possible with PRIMO to closely match simulation results with a particular linac, the results obtained with the default beam parameters are consistent with the typical values found for 6 MV photon beams from Varian Clinac 2100 linacs.

28

Field size

N linacs Parameter IROC

Median±SD PRIMO Value±SD PRIMO-IROC (%)

10×10 116

Dmax depth (cm) 1.499±0.009 1.527±0.018 1.9 PDD z=5 cm 1.295±0.005 1.310±0.015 1.1 PDD z=15 cm 0.761±0.003 0.757±0.010 -0.5 PDD z=20 cm 0.576±0.003 0.568±0.080 -1.3

6×6 99

PDD z=5 cm 1.327±0.005 1.340±0.016 1.0 PDD z=15 cm 0.746±0.003 0.739±0.090 -1.0 PDD z=20 cm 0.556±0.004 0.545±0.070 -2.0

20×20 99

PDD z=5 cm 1.254±0.003 1.257±0.013 0.2 PDD z=15 cm 0.784±0.003 0.781±0.090 -0.4 PDD z=20 cm 0.609±0.003 0.603+0.070 -1.0

6×6

104 Output factor at Dmax

0.957±0.006 0.969±0.015 1.2 15×15 1.033±0.006 1.021±0.013 -1.2 20×20 1.055±0.007 1.040±0.015 -1.4 30×30 1.082±0.008 1.063±0.014 -1.7

40×40 116 Off-axis ratio 5 cm left 1.029±0.008 1.014±0.013 -1.5 Off-axis ratio 10 cm avg 1.041±0.008 1.029+0.014 -1.2 Off-axis ratio 15 cm left 1.053±0.012 1.049±0.015 -0.4

2×2

4 IMRT-style MLC output factor at z=10 cm

0.808±0.004 0.820±0.008 1.5 3×3 0.853±0.001 0.860±0.090 0.9 4×4 0.889±0.005 0.895±0.010 0.6 6×6 0.940+0.002 0.939±0.010 -0.1 2×2

21 SBRT-style MLC output factor at z=10 cm

0.778±0.004 0.799±0.080 2.8 3×3 0.820±0.004 0.844±0.090 2.9 4×4 0.855+0.004 0.871±0.090 1.8 6×6 0.915±0.003 0.930±0.010 1.7

Conclusions The PRIMO default initial beam parameters for 6 MV photon beams from Varian Clinac 2100 linacs allows obtaining dose distributions in a water phantom which agree within 3% with a database of dosimetric data based on measurements on a large series of linacs of the same model. The findings of this work represent a first step in the validation of PRIMO for independent verification of radiotherapy plans computed by a commercial treatment planning system.

Conflicts of interest: None

References

1 D.S. Followill et al., “The Radiological Physics Center’s standard dataset for small field size output factors,” J. Appl. Clin. Med. Phys. 13(5), 282–289 (2012).

29

2 J.R. Kerns et al., “Technical Report: Reference photon dosimetry data for Varian accelerators based on IROC-Houston site visit data,” Med. Phys. 43(5), 2374–2386 (2016).

30

A Monte Carlo study of Varian Unique linear accelerator using PRIMO software

Sinan Irmak

Acıbadem Kayseri Hospital, Department of Radiation Oncology, Kayseri, Turkey Presenting author: irmakster@gmail.com Purpose A Varian Unique linear accelerator which has only 6 MV photon energy is simulated by using Primo software for dosimetric quality control purposes. Material and Methods Primo version version 0.1.5.1307 is used. For S1 section default parameters are used for primary beam parameters. Splitting roulette VRT technique is used while size of splitting region is fitted to the field size of 27 cm x 27 cm that set in S2. S1 is simulated only once and it took 24.5 hours to simulate 2.04x107 histories with a 12 core CPU. For S2 section field sizes of 5cm x 5cm, 8cm x 8cm, 10cm x 10 cm, 15cm x 15cm and 20 cmx 20 cm square fields are simulated by using all the histories acquired from the simulation of S1. To investigate the dose deposition a water phantom is used in S3 section. Splitting factor is tuned to obtain uncertainties below 4.2%. Results PDD curves for first four fields are compared with experimental data obtained by three different PTW dedectors. Integral ratio is chosen as normalization mode and all curves found in agreement better than 99% when 2%/2mm passing criteria is used for gamma analysis. Agreement between Pinpoint dedector is slightly better than the 0.125 cm3 Semiflex ion chamber and Diode E semiconductor dedector for the field sizes smaller than 10cm x 10cm. Crossplane profiles are also compared with the data measured with 0.125 cm3 Semiflex ion chamber in 5cm depth and 10cm depth respectively for field sizes up to 20 cm x 20 cm. While the percentage of the points passing the criteria of 3%/3mm are found more than 97.6% for fields smaller than 15 cm x 15 cm in both depths, for 20x20 cm fields percentage of points passing this criteria is found below 85%. Conclusions Primo is a very convenient tool to simulate medical linear accelerators. Without any coding effort, one can obtain good dosimetric results to compare with the experimental data. In order to have realistic results especially for field sizes bigger than 15 cm x 15 cm 2x107 histories are inadequate.

31

Monte-Carlo simulation of 3D-scanner in PRIMO: comparison with experimental data

R. Tortosa1, C. Senra2, S. Díez2, R.M. Cibrian3

(1) Department of Radiotherapy Oncology, Hospital IMED Elche, C/ Max Planck 3—Parque Empresarial, Elche (Alicante), Spain. (2) Department of Radiophysics, Hospital Clínico Universitario, Av. Blasco Ibañez 17, Valencia, Spain (3) Department of Physiology, Universitat de Valencia, Av. Blasco Ibañez 13, Valencia, Spain. (*) Presenting author: rtortosa@imedhospitales.com Objective The objective of the presenting work is to introduce a cylindrical water phantom in PRIMO software and compare the simulated results with those experimentally performed in a commercial cylindrical water phantom. Material / Methods The cylindrical water phantom was generated using Matlab R2010 with 69-cm of diameter, 102-cm of height and bin size of 0,4 x 0,5 x 0,4 cm. Then it was imported to PRIMO v1.5.1307. In order to perform the simulations, Varian2300 model from the list of clinacs was selected for photon energy of 6 and 15 MV respectively. Initial parameters used in each simulation were 5,4 MeV for initial energy of 6 MV and 14.4 MeV for 15 MV respectively, 0,05 for energy FWHM (full width at half maximum), 0,1 cm for focal spot and 1º for divergence. The number of total histories simulated were 150·106 for 6 MV and 27·106 for 15 MV. Experimental measurements were performed in a Varian Clinac DHX (Varian, Palo Alto, CA) with photon energy of 6 and 15 MV. The water phantom used was the 3D-Scanner (Sun Nuclear, Melboume, FL) which is the only cylindrical water phantom commercially available Dose profiles and percentage depth doses were measured for field sizes of 40x40, 30x30, 20x20, 10x10 y 4x4 cm2 for 6 and 15 MV respectively. Selected depths to compare results were done at maximum of each energy (1,4 cm for 6 MV and 2,7 for 15 MV), 10 cm and 20 cm. All simulations were performed at 100 cm surface-source distance. Results Gamma analysis with the criteria of 3%-3mm, 2%-2mm and 1%-1mm was performed among the measurements and the simulations. Percentage depth dose results show in all cases and criteria values higher than 95,3 % so simulated energy is equivalent to experimental energy. In Figure 1a and 1b, all results were plotted in box diagrams for each energy. The highest variability appears in 1%-1mm criteria. Points passing the criteria decrease with field size and depth. This effect can be explained due to bin size

32

because PRIMO software doesn’t allow bin sizes lower than those used in this work. The highest statistic uncertainty value in all simulated cases was 1,1%.

Figure 1: Box diagram for 6 MV (1a); box diagram for 15 MV (1b) Conclusion PRIMO allows introducing phantoms of different sizes and geometries to modelate radiation beam in some commercial clinical accelerators. A list of commercial water phantoms could be added in future versions of Primo to provide users extra simulating tools.

33

Monte Carlo simulations in radiotherapy dosimetry

Pedro Andreo Department of Medical Radiation Physics and Nuclear Medicine, Karolinska University Hospital, and Department of Oncology-Pathology, Karolinska Institutet, SE-171 76 Stockholm, Sweden. Presenting author: pedro.andreo@ki.se The use of the Monte Carlo (MC) method in radiotherapy dosimetry is considered to be pioneered by Berger and Seltzer in the late 1960s with the calculation of electron depth-dose and fluence distributions in water, followed by their calculations of stopping-power ratios (SPRs) some years later. The 50 years of development and considerable increase in computer speed following those early works have made the technique a fundamental tool in the field, which nowadays includes the simulation of radiotherapy treatment machines, detectors and dose planning calculations. Recall the basic expressions and dosimetric key quantities for radiotherapy dosimetry. SPRs and energy-absorption coefficients ratios are derived quantities that rely on basic quantities, such as mass stopping powers and µen/r values, updated by ICRU Report 90 (2016), and on electron and photon fluences. The latter form the grounds of MC calculations in dosimetry, used to calculate SPRs and µen-ratios, which subsequently enter into the small and large detectors cavity theories. All derived key quantities for clinical dosimetry are currently based on MC simulations. As is well-known, Bragg-Gray (BG) theory relies on the assumption that the electron fluence in the medium is not perturbed by the presence of a detector (i.e., FmedºFdet, yielding charged-particle equilibrium, CPE), but the detector does perturb the electron fluence. Classically, the problem was “solved” introducing a detector perturbation correction factor, pdet, and assuming that the approximation FmedºFdet still holds. The last 15 years have witnessed numerous MC simulations of pdet. The state of the art is that phase-space data for specific machines are used to simulate in detail the response, inside a phantom, of a given detector whose geometry is described accurately. There are, however, a number of issues emerged in this area, such as: (i) the accuracy in describing detector geometries and/or MC transport parameters, which has resulted in rather different pdet for a given detector, even using the same MC code; (ii) experimentally demonstrated detector-to-detector differences of a given type, which cannot be reproduced in simulations based on “ideal” manufacturer drawings; and (iii) the applicability of BG theory, which fails in small beams. MC calculations do not require CPE, but our current formulations rely on CPE-based expressions. The major advantage in using the classical solution for BG theory is that conventional SPRs based on the assumption FmedºFdet can be used. In parallel, a major constrain is that pdet for the different detector components must be small and independent of each other. In small MV fields,

34

however, pdet factors for many detectors can be up to 10% and often CPE is lacking. This means that the BG assumptions used so far break down. A solution, now generally adopted, is to compute directly the ratio between the dose to a point in water and the mean dose inside the detector. The procedure does not rely on BG equations and/or CPE; it yields a generic dose-conversion factor that includes all possible entanglements between detector-specific components and their influence on Fdet. The conclusion is that the assumptions of FmedºFdet and small independent pdet are no longer needed and should not be used in dosimetry (except for pedagogic purposes). MC in radiotherapy treatment planning (MCTP) has only been possible since recent years; unfortunately, for sake of speed, most commercial MCTPs are “trimmed” for low-Z media. They are claimed to yield results within the requirements for Treatment Planning Systems, even in the presence of inhomogeneities. There are however three important questions on MCTP that deserve attention: i) should dose-to-tissue (Dtis) or dose-to-water (Dw) be calculated?; ii) does MCTP inherently calculate Dtis accurately?; and (iii) how should the conversion between Dtis and Dw be performed?. These issues will be discussed in detail in the lecture.

35

PENELOPE transport parameters for linac simulation

Josep Sempau*,1, Miguel Rodriguez2, Lorenzo Brualla3

(1) Institut de Tècniques Energètiques. Universitat Politècnica de Catalunya. Diagonal 647. 08028 Barcelona. Spain. (2) Centro Médico Paitilla. Ciudad de Panamá. Panama. (3) NCTeam. Strahlenklinik. Universitätsklinikum Essen. Hufelandstr 55. 45147 Essen. Germany. (*) Presenting author: josep.sempau@upc.es The simulation of high-energy electron transport requires the use of condensed history methods in order to reduce the computation time to reasonable values. These methods, consisting in describing the collective effect of a large number of interactions occurring along an electron step as a single artificial event, is known to produce artifacts when particles move across or in the vicinity of heterogeneities [1]. Artifacts, which manifest as a dependence of results on the step size, become prominent when one of the regions adjacent to an interface has a mass thickness orders of magnitude smaller than that of the neighbor region. This situation is typically found in the simulation of small ion chambers [2], one of the challenges of the Monte Carlo simulation of charged particles. With the exception of the air in the monitoring ion chamber, a linac geometry does not contain dramatic variations in mass thickness. Therefore, except when the quantity of interest is the chamber signal, in principle one does not expect simulation results to be very sensitive to variations of the step length. This assumption contradicts some results found by our group [3], which showed that the selection of the transport parameters that control the step length in the linac target is crucial to avoid heavily biased dose distributions in the patient. Deviations in the central axis as high as 17% where found with respect to a case where condensed simulation was not used. The problem was attributed to the fact that small changes in the electron angular scattering distribution propagate to the angular distribution of emitted bremsstrahlung photons. These photons subsequently travel about 1 m before entering the patient, where the dose distribution is analyzed at the mm level. Thus, angular deviations in the order of 1mm/1m, that is, 1 mrad, which are irrelevant in most common situations, become relevant. The artifact is not trivial to detect because the dose bias can be masked by using unrealistic primary beam data. Thus, certain combinations of primary beam energy and FWHM of the focal spot size produce correct dose profiles despite the inadequate choice of the transport parameters in the target. The application of some variance reduction techniques may also play an indirect role, introducing further confusion into the matter and making more difficult to isolate the root of the problem. PENELOPE's model, on which PRIMO relies, defines eight transport parameters for each material. Of these, only two (called C1 and C2) directly affect the electron

36

angular distribution. The variation of the dose with C1&C2 was studied in detail and a recommendation of acceptable values was issued. The potential relevance of this problem for other MC codes is briefly commented. References [1] A F Bielajew, D W O Rogers and A Nahum, Phys. Med. Biol. 30 (1985) 419 [2] J Sempau and P Andreo, Phys. Med. Biol. 51 (2006) 3533 [3] Miguel Rodriguez, Josep Sempau and Lorenzo Brualla, Med. Phys. 42 (2015) 2877

37

Monte Carlo simulations of intensity modulated radiotherapy using PRIMO software

Alessandro Esposito*,1,2, Jorge Oliveira2, Sofia Silva2,3,

Joana Lencart2,3, Joao Santos2,3 (1) Azienda Ospedaliera Santa Maria, Radiotherapy Department, Terni, Italy. (2) Medical Physics, Radiobiology and Radiation Protection Group, IPO Porto Research Center (CI-IPOP), Portuguese Oncology Institute of Porto (IPO Porto), Porto, Portugal. (3) Medical Physics Department, Portuguese Oncology Institute of Porto (IPO Porto), Porto, Portugal. (*) Presenting author: alessandro.espo1980@gmail.com Purpose/Objective Intensity Modulated Radiation Therapy (IMRT) allows creating complex dose distributions. The dose distribution is calculated using different algorithms, but Monte Carlo (MC) is the gold standard for its complete description of radiation-matter interaction [1]. MC can be a powerful tool for dosimetric plan evaluation, Quality Assurance (QA) programs and treatment quality improvement. Several MC codes are in use for Radiotherapy applications. Recently, PRIMO was developed [2] with a user-friendly interface. Nevertheless, IMRT is not introduced yet. The aim of this work is to present an algorithm to configure the PRIMO to simulate IMRT and to show the results of a preliminary study. Materials and methods Two Radiotherapy units with the same Varian 2300CD LINAC head were considered. One unit uses Millennium120 and the other has 120HD as Multi Leaf Collimator (MLC). The primary beam calibration was performed on the basis of experimental measurements (PDD and dose profiles) in water tank. The MLCs in use on both the units were further validated simulating static irradiations and comparing the results with measurements using Gafchromic films. Dedicated software was developed to allow automatic configuration of PRIMO to simulate MLC motion by randomly sampling static MLC configurations. A preliminary test was performed on a simple leaf motion to check the algorithm and a second test simulated a complex MLC motion in solid water. Comparisons with experimental was performed by creating simulated dose images at specific planes by automatic software manipulation of the PRIMO output files. Assessments were performed using the 2D Gamma analysis (2%, 2mm). Calculation time was studied in the view of treatment simulation optimization. Results The primary beam parameters calibration showed more than 95% of Gamma points < 1 for both the units as shown in Figure 1. Simulation of static MLC configuration showed 100.0% and 99.1% of 2D Gamma points < 1, while simulations of the basic dynamic procedures resulted in 98.9% and 99.5% of Gamma points < 1 with 120HD and Millennium120 respectively. The higher

38

modulation test showed 99.1% of Gamma points < 1 with respect to the experimental using the 120HD as shown in Figure 2. Increment of the agreement between IMRT simulation and Gafchromic measurement was observer when the number of static configurations to reproduce the MLC motion is incremented. The same trend is observed for the total calculation time. Conclusion A workflow was found to drive PRIMO to simulate IMRT procedures. The results of simulations agree with the experimental measurements, indicating PRIMO software as a potential tool for clinical implementation of MC simulations of dynamic techniques such as IMRT. The agreement increment between simulation and measurements with higher number of static fields was obvious, but the total calculation time increment was unexpected and addressed to the PRIMO post-processing algorithm. Conflict of interests The authors have no relevant financial or non-financial relationship to disclose. References [1] Reynaert, N., et al.: Monte Carlo treatment planning for photon and electron beams. Rad Phys Chem (2007), 76, 643-686 [2] Rodriguez, L., et al.: PRIMO: A graphical environment for the Monte Carlo simulation of Varian and Elekta linacs. Strahlentherapie und Onkologie (2013) 189, 881-886

39

FIGURE 1: Experimental (red) and simulated (blue) PDD (top), X-profiles (left bottom) and Y-profiles (right bottom). The green data represent the Gamma values reported according to the right vertical axes.

FIGURE 2: Dose distribution comparison between experimental data as acquired by the Gafchromic film (top left) and the simulated result using 100 random static fields (top right) for the high modulation dynamic delivery described. The 2%, 2mm evaluation (left bottom) showed 99.1% of gamma points lower than 1. PTW Verisoft was used to calculate the gamma values. At bottom right the dose distribution at the film location when only 20 fields are used.

40

The Dose Planning Method code

Josep Sempau Institut de Tècniques Energètiques. Universitat Politècnica de Catalunya. Diagonal 647. 08028 Barcelona. Spain. Presenting author: josep.sempau@upc.es The Dose Planning Method (DPM) code was conceived in the early 2000's [1] as a fast Monte Carlo engine for photon and electron transport in external beam radiotherapy applications, with optimized algorithms for voxelized geometries. The limited range of atomic numbers and beam energies needed for radiotherapy patients allowed a substantial simplification of the physics models. This, combined with a novel approach for the treatment of electron multiple scattering and of the transport across voxel heterogeneities, resulted in a code that is nearly as accurate as most general-purpose Monte Carlo codes but around one order of magnitude faster. The DPM code, which is free and open source, can be downloaded from http://www.upc.es/inte/downloads. In this talk the basic principles of the algorithm will be succinctly presented to provide a solid background for the presentation on the improvements introduced in PRIMO that will be given later by M. Rodríguez, one of the PRIMO coauthors. The discussion of the physics, albeit very brief, will provide some perspective on the main differences between PENELOPE and the DPM algorithms, which is relevant because PRIMO will allow for a direct comparison between the two systems. Photon cross section models will be outlined and the transport in a borderless-like geometry using the so-called Woodcock tracking, discussed. In fact, the DPM development effort focused mainly on the electron transport algorithm, which is a more time-consuming task than that of photons. Electron cross section models are therefore also discussed, with emphasis on the novel approach mentioned above for elastic scattering and inter-voxel transport. In this context, the multiple scattering formalism used, a refinement of that from [2], is briefly introduced. The DPM code was extensively benchmarked in the early 2000's by a mixed group from the Univ. of Michigan (UoM) lead by I. J. Chetty and A. F. Bielajew (see, e.g., [3]). Eventually, DPM was integrated into UoM's in-house treatment planning system (TPS) called UMPlan. Also, version 3 beta of the ADAC Pinnacle TPS was based on a C++ port of DPM. ADAC was acquired by Philips Medical Systems in 2000. Philips announced, but never released, the Pinnacle version based on DPM. Other companies and research groups have adapted DPM to act as the Monte Carlo engine of their applications (TPSs, internal-emitter therapy, intraoperatory radiation therapy, etc.), some of them after parallelizing or vectorizing the code.

41

An UoM effort [4] to combine linac and patient simulation, in a way similar to PRIMO, based on the BEAM/EGS (for the linac head) and DPM (for the patient) will be commented. References [1] J. Sempau, S. J. Wilderman & A. F. Bielajew, Phys Med Biol 45 (2000) 2263 [2] I. Kawrakow & A. F. Bielajew, NIMB 134 (1998) 325 [3] I. J. Chetty et al., Med. Phys. 29 (2002) 1837 [4] N. Tyagi et al., 44th AAPM annual meeting, 2002

42

Improvements on DPM for PRIMO

Miguel Rodríguez

Centro Médico Paitilla, Ciudad de Panamá. Panama Presenting author: milrocas@gmail.com PRIMO is conceived with a layer structure in which the upper layer is the Graphical User Interface. Lower layers encompass the Monte Carlo radiation transport code and the geometries (linacs and patient). The communication between layers is accomplished by sharing input/output files. So far, the only one Monte Carlo simulation code used in PRIMO is the general-purpose code PENELOPE [1]. PENELOPE is composed of a set of subroutines that must be driven by a main program to describe the sources of particles, to accomplish the radiation transport in a given geometry and to produce tallies. For such a purpose, PRIMO uses the penEasy program [3]. PENELOPE incorporates a very accurate physics and its geometry model is quite general, hence it can be applied to diverse research problems. However, typical simulation times obtained with PENELOPE are not practical for routine dose calculations in the field of radiotherapy. The Dose Planning Method (DPM) [3] is a fast MC code particularly oriented to dose calculation in radiotherapy. Despite incorporating a series of approximations, DPM has proven to agree within 1-2% with measurements in a variety of conditions [4, 5]. Furthermore, DPM has the advantages of its relatively small code compared to a general-purpose MC code and the fact that it is open source which makes more feasible to introduce changes. In this talk we introduce the improvements made to the original version of DPM aimed at to incorporate it as an additional radiation transport code in PRIMO. The main improvements made to DPM are:

• The use of phase-spaces as source of particles. • The parallelization of the code for many- and multi-core computer

architectures. • The radiation transport in the linac geometry.

Result of benchmarks that emphasize the differences between DPM and PENELOPE are presented. DPM is expected to be less accurate for simulations in no “water-like” materials. Despite using the PENELOPE subroutines to move the particle in the linac, the physics of the transport implemented in the original DPM was not changed in the improved version. Therefore, the discussion is centered on what deviations are produced in the absorbed dose by the transport in high-density materials presented in the patient-dependent part of the linac.

43

References [1] F Salvat, J M Fernández-Varea, and J Sempau. (OECD Nuclear Energy Agency, Issy-les-Moulineaux, France, 2011). [2] J Sempau, A Badal, and L Brualla. Med. Phys. 38, (2011) 5887–5895. [3] J Sempau, S J Wilderman and A F Bielajew. Phys. Med. Biol. 45, (2000) 2263–2291. [4] I Chetty et al Med. Phys. 29 (2002). [5] I Chetty et al Med. Phys. 30 (2003).

44

What’s new in PRIMO

Miguel Rodríguez*,1, Josep Sempau2, Lorenzo Brualla3

(1) Centro Médico Paitilla, Ciudad de Panamá. Panama. (2) Institut de Tècniques Energètiques. Universitat Politècnica de Catalunya. Diagonal 647. 08028 Barcelona. Spain. (3) NCTeam. Strahlenklinik. Universitätsklinikum Essen. Hufelandstr 55. 45147 Essen. Germany. (*) Presenting author: milrocas@gmail.com Almost two years have elapsed since the last distribution of a PRIMO version. During that time substantial improvements have been done, the most relevant are:

• A parallel version of the Dose Planning Method (DPM) fast Monte Carlo code has been added to PRIMO. DPM is particularly oriented to dose calculation in radiotherapy.

• Treatment plans can be imported from Treatment Planning Systems provided they are in DICOM format.

• Dose calculation in dynamic treatments such as IMRT and VMAT can be simulated with DPM as well as with PENELOPE.

• New analysis functions have been incorporated such as those to compare 3D dose distributions.

• Many changes have been introduced in the user interface and graphic visualization has been improved through the use of OpenGL.

Along this talk, changes in PRIMO are explored in detail and indications are given aimed at to take full advantage of the new capabilities implemented in this system.

45

Study of build-up dose region for 6MV photon beam using PRIMO simulation software

Mamta Mahur*,1,3, PS Negi1, Manoj Semal2, Munendra Singh3

(1) Delhi State Cancer Institute, Dilshad Garden, Delhi. India. (2) Army Hospital (R&R), Delhi. India. (3) Sharda University, Greater Noida. India. (*) Presenting author: mamtamahur@gmail.com Objective This study aims to the Monte Carlo simulation of Varian 600C Linear accelerator using PRIMO simulation software and do the comparative analysis of simulated depth dose curves in buildup region with the measured depth dose curves. Materials and Methods PRIMO simulation software (version 0.1.5.1307) which is PENELOPE based Monte Carlo software was used to simulate PDD (depth dose) curves for Varian 600C (Varian Medical Systems, Palo Alto, CA, USA) medical linear accelerator for 6MV photon Energy in homogeneous water phantom for field sizes 5x5cm2, 10x10cm2, 15x5cm2 at SSD100cm. Linac was simulated with the preset structural details of linear accelerator from wish list available in PRIMO software. For 6MV photon beam nominal energy was set to 5.4Mev, focal spot was set to zero. Measurements for PDD (depth dose curves) were performed using RFA300 computer controlled scanning water phantom and RK ionization chamber for field sizes 5x5cm2, 10x10cm2, 15x5cm2 at SSD 100cm. To evaluate the simulation each simulated PDD curve was compared with the measured curve in dose analysis part of software. Doses were compared in buildup dose region for the three field sizes as analysis of depth doses in buildup region have added uncertainty for most of the treatment planning calculation algorithms. Results Comparative analysis of simulated depth dose curves with the measured PDD shows good agreement. With increase in number of histories the observed difference between simulated and measured depth dose curves was found to be decreased significantly. Conclusion PRIMO is a user-friendly Monte Carlo simulation software which can be used for verification of clinical photon and electron beams in dose buildup region.

46

Determination of initial parameters of Varian 2100 CD linac electron beams using PRIMO

Reza Maskani*,1,2 Mohammad Javad Tahmasebibirgani3,4,

Mojtaba Hoseini-Ghahfarokhi2,3, Jafar Fatahiasl2,5

(1) School of Allied Medical Sciences, Shahroud University of Medical Sciences, Shahroud, Iran. (2) Student Research Committee, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran. (3) Department of Medical Physics, School of Medicine, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran. (4) Departments of Clinical Oncology, Golestan Hospital, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran. (5) Department of Radiology, School of Paramedicine, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran. (*) Presenting author: maskany@gmail.com Purpose Finding out primary characteristics of electron beams for a Varian 2100C/D linear accelerator by recently developed Monte Carlo software; PRIMO and to verify relations between electron energy and dose distribution. Material and methods To have conformity of simulated and measured dose curves within 1%/1mm, mean energy, full width at half maximum (FWHM) of energy and focal spot FWHM of initial beam were changed iteratively. Transport parameters were: C1=C2=0.1, WCC=100 KeV and WCR=20 KeV. Cutoff energies for electrons and photons were EABS(e)=100KeV and EABS (ph)=20 KeV. Two well-known empirical relations between E0 and R50 are (Eq1) and (Eq2). E0 and R50 are in MeV and cm, respectively. Mean energies were extracted from validated phase spaces and compared with related equation. To explain the importance of correct estimation of primary energy on clinical case, computed tomography images of a thorax phantom were imported in PRIMO. R50 of 6 MeV electrons was shifted intentionally as much as 2mm downward. Then, corresponding initial beam energy was found by matching calculated and artificial measured R50. Dose distributions and dose volume histogram (DVH) curves were compared between confirmed case and an artificial case with overestimated energy. Results After frequent changes in the initial parameters and performing the simulations, maximum conformity between measured and simulated curves were obtained by using values are listed in Table 1. Comparison of measured and simulated PDD curves is shown in Fig1.

47

Table 1: Validated initial beam parameters for different nominal energy

Nominal Energy(MeV)

Initial energy(MeV)

Energy FWHM(MeV)

Focal spot FWHM(cm)

6 6.68 1.2 0.4 9 9.73 0.8 0.3 12 13.2 0.6 0.4 15 16.4 0.3 0.4

Fig1: Comparison of measured and simulated PDD curves for two different nominal energy. Results pointed out energy FWHM reduced with increase in energies. Three mm focal spot FWHM for 9 MeV and 4 mm for other energies make proper match among simulated and measured profiles. In addition, the maximum difference of calculated mean electrons energy at the phantom surface with empirical equations was 2.2 percent (Table 2). Finally, it is shown there are clear differences in DVH curves of validated and artificial energy as heterogeneity index was 0.15 for 7.21 MeV and 0.25 for 6.68 MeV (Fig2). Table 2: mean energy (E0) of simulated energy spectra at the phantom surface comparing with empirical equations results.

Nominal Energy (MeV)

E0 (MeV) MC Calculated

Equation 1

Equation 2

6 5.77 5.5 5.64 9 8.34 8.37 8.33 12 11.53 11.67 11.52 15 14.28 14.68 14.5

48

Fig2: DVHs for CTVs and Right Lung as organ at risk in both Artificial (7.21 MeV) and Real (6.68 MeV) energy. Conclusion Monte Carlo model presented in PRIMO for Varian 2100 CD was precisely validated with our experimental data. IAEA polynomial equation estimated mean energy more accurate than known linear one. Small displacement of R50 changes DVH curves and homogeneity indexes. Therefore, R50 should be carefully measured to avoid errors in determination of initial energy. PRIMO is user-friendly software which has suitable capabilities to calculate dose distribution in water phantoms or computerized tomographic volumes accurately.

49

Dosimetric evaluation of AAA and Acuros XB algorithms with the PRIMO software in inhomogeneous media

Martín Albina, ML.1; Pizarro Trigo, F.1; Jiménez Albericio, J.*,2;

Nuñez Martínez, LMR.1; Morillas Ruiz, J.1; Sánchez Jiménez, J.1

(1) Hospital Universitario de Burgos, Burgos, Spain. (2) Hospital Universitario Miguel Servet, Zaragoza, Spain. (*) Presenting author: javierjimenez.a@gmail.com Purpose/objective The purpose of this study is to investigate the behaviour of the Analytical Anisotropic Algorithm (AAA) and Acuros XB (AXB) algorithms implemented in the Eclipse treatment planning system (Varian Medical Systems) when calculating absorbed doses in inhomogeneous media and to compare it against a fully Monte Carlo simulation using the PRIMO software. Materials and methods The study is carried out in a virtual phantom characterised by simple geometrical structures. A 30x30x15.6cm3 water phantom is designed with layers of different materials and densities in the following order from surface: a 5cm water layer (1g/cm3), a 1cm bone layer (1.853g/cm3), a 1.6cm air layer (0.0012g/cm3), a 2cm adipose tissue layer (0.92g/cm3) and a 1cm cartilage layer (1.42g/cm3). 6 MV photon beams, a 5x5cm2 field size, a 0º gantry and collimator angle and a 100cm source-surface distance are employed. With regard to the calculations with Eclipse, a Varian Clinac DHX is implemented in the treatment planning system. A 0.1cm dose matrix size is used for both AAA and AXB algorithms (version 11.03.31). The Dose to medium calculation option is chosen for AXB. Concerning the PRIMO system (version 0.1.5.1300, based on the codes PENELOPE 2011), the modeled geometry of a Varian 2300 C/D is used for the simulations. About 3.108 histories are simulated. The employed variance-reduction techniques are Splitting-Roulette for phase-space s1 and a splitting factor of 125 for s3. For s3 the chosen bin size is x=0.3cm, y=0.3cm, z=0.1cm. Two simulations are done: the first one, using PRIMO’s default value of the parameters Eabs, C1, C2, Wcc, Wcr and smax (Table 1); the second one, assigning the “safe” values recommended by the Penelope-2014 manual to the simulation parameters (Table 2). In each case, percentage depth dose (PDD) curves are obtained and compared. The uncertainty of the result, given at two standard deviations, is 1.3%.

50

Results The results are shown in Figures 1 and 2. The PDD curves obtained with AAA and AXB are represented in both images. The resulting PDD curves from the first and second PRIMO simulations are added to Figures 1 and 2 respectively.

Figure 1 Figure 2 As can be seen in Figure 1, the PRIMO PDD result is not correct in the air layer, where no energy deposition at all is represented. However, in Figure 2, a more feasible PRIMO PDD curve is obtained, showing good agreement with other Monte Carlo results taken from literature references, therefore it is assumed as the best approach. By analysing the PDD curves in Figure 2, a very good agreement is obtained between the PRIMO software and AXB even in extreme cases of very low density materials whereas the AAA curve shows larger differences in these situations. Conclusions AXB is a more suitable algorithm than AAA when calculating absorbed doses in regions of very low densities. A proper selection of the Eabs, C1, C2, Wcc, Wcr and smax parametres is necessary in order to better obtain the most accurate simulation results.

51

References 1. Fogliata et al. Dosimetric evaluation of Acuros XB Advanced Dose Calculation algorithm in heterogeneous media. Radiation Oncology 2011 6:82 2. Han et al. Dosimetric comparison of Acuros XB deterministic radiation transport method with Monte Carlo and model-based convolution methods in heterogeneous media. Med.Phys.Vol.38 No.5 (2011)

52

A Monte Carlo study for a breast conformal radiotherapy plan using PRIMO software

Sinan Irmak

Acıbadem Kayseri Hospital, Department of Radiation Oncology. Kayseri. Turkey. Email: irmakster@gmail.com Poster presentation Purpose A left sided whole breast radiotherapy plan is simulated by using PRIMO software for comparing the dose distributions calculated by Eclipse treatment planning system. Material and Methods PRIMO version version 0.1.5.1307 is used to simulate a Varian Unique medical linear accelerator. S1 is simulated by using splitting roulette VRT technique while size of splitting region is fitted to the field size of 26 cm x 26 cm that set in S2. S1 is simulated only once and it took 60 hours to simulate 5.04x107 histories with a PC equipped with a Intel Xeon E5-2620 12 core CPU and 32 GB of RAM. Water phantom simulations has been done for square fields from 3 cm x 3 cm to 20 cm x 20 cm and splitting factor is tuned to obtain uncertainties below 4.12%. Every PDD curve found in agreement better than 99% when 2%/2mm passing criteria is used for gamma analysis and lateral dose profiles for the same field sizes in depths of 5 cm and 10 cm found in agreement better than 95% when 3%/3mm passing criteria is used for gamma analysis when they compared to experimental data measured with PTW 31010 0.125 cc ion chamber. After the linac model is validated a treatment plan is applied to CT image for comparing the relative dose results given by Varian Eclipse and PRIMO for a left sided early stage whole breast radiotherapy case. 3-dimensional conformal radiotherapy with field-in-field technique is used. Same CT calibration curve in Eclipse used in PRIMO. Two tangential fields with two in field beams are used. MLCs are positioned as same positions with treatment planned by using Eclipse for four fields. AAA 13.6 dose calculation algorithms are used to calculate relative dose deposition. Average uncertainty is found 3.695% after dose calculation in CT medium. No plan normalization is made. Results Relative doses to the normal tissues and the target are found as follows in Eclipse and PRIMO.

53

PRIMO Eclipse Heart V%10 36.78 22.96 Left Lung V%10 25.03 33.97 Left Lung V%40 23.00 23.18 Right Breast V%10 2.69 0.37 Lungs V%40 10.58 10.63 CTV V%80 82.48 100 CTV D%95 77 96.6

Conclusions PRIMO is a very convenient tool to simulate medical linear accelerators without any coding effort. But in order to compare dosimetric calculation results with commercial treatment planning systems history number in order of 107 may be inadequate. Imprecise adjustment of the same isocenter with the Eclipse plan and calculation methods for tangential field dose calculations may have caused such differences in doses for targets and normal tissues in this study.