ESTRO 38 Abstract book

S1023 ESTRO 38

EP-1884 Commissioning and clinical validation of FRED: Monte Carlo on GPU for proton beam therapy M. Garbacz 1 , J. Baran 1 , G. Battistoni 2 , M. Durante 3 , J. Gajewski 1 , N. Krah 4 , K. Krzempek 1 , V. Patera 5 , M. Pawlik- Niedzwiecka 1 , I. Rinaldi 6 , B. Sas-Korczynska 7 , E. Scifoni 3 , A. Skrzypek 8 , A. Schiavi 5 , F. Tommasino 9 , A. Rucinski 1 1 Institute of Nuclear Physics PAN, Proton Radiotherapy Group, Krakow, Poland ; 2 INFN, Sezione di Milano, Milan, Italy ; 3 TIFPA, Trento Institute for Fundamental Physics and Applications, Trento, Italy ; 4 CNRS/CREATIS, Umr 5220, Lyon, France ; 5 Sapienza Università di Roma, Dipartimento di Scienze di Base e Applicate per Ingegneria, Rome, Italy ; 6 ZonPCT/Maastro Clinic, ZonPCT/Maastro Clinic, Maastricht, The Netherlands ; 7 Maria Sklodowska-Curie Institute - Oncology Center, Oncology Clinic, Krakow, Poland ; 8 AGH University of Science and Technology, Faculty of Physics and Applied Computer Science, Krakow, Poland ; 9 University of Trento, Department of Physics, Trento, Italy Purpose or Objective This work presents commissioning and validation of GPU- accelerated Monte Carlo (MC) code FRED at proton beam therapy (PBT) facility in Krakow. The aim of this study was to develop fast proton beam model phase space implementation method exploiting experimental data acquired during the facility start-up. We validate FRED simulations against patient quality assurance (QA) measurements following clinical acceptance procedures. Ultimately we utilize time performance of FRED for recalculation of clinical treatment plans with variable radiobiological effectiveness. Material and Methods A GPU-accelerated MC tool FRED (Fast paRticle thErapy Dose evaluator; Schiavi et al. 2017) was developed at Sapienza University of Rome and is investigated at Krakow PBT centre. FRED was already validated against FLUKA and GEANT4, as well as clinical data from CNAO facility. The depth-dose distributions (DDDs) of single pencil beams measured in water phantom and lateral beam profiles in air for 17 energies in 70-226.1 MeV range were used to build a FRED dedicated phase space library. The validation procedures included QA measurements of spread out Bragg peak (SOBP) of different range in water and comparison of more than 182 measurements of Krakow patient treatment fields. The gamma index (GI) analysis were used to evaluate the dosimetric results of MC calculations. Results The DDDs of a single proton pencil beam in water for various energies simulated in FRED MC code were in agreement with the commissioning measurements: the range (R80%) of the pencil beams agreed within 0.1 mm, the absolute dose difference along the pencil beam profile was below 5%, the FWHM of the Bragg peak agreed within 0.3 mm, the distal fall-off width between 80% and 20% Bragg peak dose agreed within 0.07 mm. The simulations of verification plans in water were performed and evaluated against measured data using GI method with 3%/2mm passing criteria (see example field in Figure). The dose distributions from FRED fulfill the Krakow facility QA acceptance criteria (passing rate > 90% for dose > 10% of maximum dose) with average passing rate 96.28(3.3)% (1 sigma). For a patient verification treatment plan the average tracking rate was 8.5(1.6)x10^6 protons / sec (1 sigma).

the better coverage of the high dose CTV, while respecting the constraints of the critical OAR (brainstem, chiasma, optic nerves). References : [1] Amichetti M., et al. Neurosurg Rev. 2009;32:403–16. EP-1883 Lung tumor target delineation: different segmentation strategies C. Gasperi 1 , S. Borghesi 2 , L. Noferini 1 , C. Sottocornola 1 , A.S. Curion 1 , F. Zenone 1 , P. Pernici 2 , R. De Majo 2 , S. Nucciarelli 2 , S. Nanni 2 , S. Bertocci 2 , P.G. Gennari 2 , A. Rampini 2 , L. Lastrucci 2 , C. Iermano 2 , E. Tucci 2 1 San Donato Hospital, Health Department Staff - Medical Physics, Arezzo, Italy ; 2 San Donato Hospital, Radiotherapy Department, Arezzo, Italy Purpose or Objective Precise definition of the target volume is one of the crucial factors in the management of Non-Small Cell Lung Cancer (NSCLC) with Stereotatic body radiotherapy (SBRT). It is widely recognized that the motion pattern of lung tumors varies greatly among patients. Therefore tumor motion should be assessed with patient specific image acquisition, to ensure adequate tumor coverage and minimizing dose to the Organs at Risk. The "gold standard" approach for defining an Internal Target Volume (ITV) is to use gross tumor volume (GTV) delineated over several phases in course of one respiratory cycle. It is a time consuming method and different Institutions have adopted several alternative techniques which compress all temporal information into one CT image set, to optimize work flow efficiency. The purpose of this study is to evaluate alternative target segmentation strategies with respect to the gold standard. Material and Methods Twenty lung cancer SBRT patients, treated on a linac with 4 mm width multileaf-collimator (MLC), were analyzed retrospectively. From the acquisition of a low-pitch helical CT scan (Untag CT) combined with a respiratory monitor system signal, four-dimensional CT (4D-CT) scans were reconstructed for each patient. GTV was delineated based on 4 single respiratory phases and on Maximum Intensity Projection (MaxIP), Minimun Intensity Projection (MinIP), Mean Intensity Projection (MeanIP) CTs and Untag CTs. Comparison was performed on Dice similarity coefficient (DSC). The relative position between the delineated target was evaluated calculating the centroid distance between volumes. Results GTVs derived from different MaxIP and MeanIP image sets were at least comparable with the single phase ITV delineation, with DSC values varying from 0.551 to 0.936 for Untag, from 0.331 to 0.877 for MaxIP, from 0.354 to 0.877 for MeanIP. MinIP GTV delineation was less comparable to the 4D-CT ITV with DSC range between 0.07 and 0.8. The differences in relative position of target volume localization were small and in all cases < 3mm. The mean differences ± SD of the centroid distances for ITV and GTV_Untag, GTV_maxIP, GTV_meanIP, GTV_minIP were 0.18±0.16 cm, 0.18±0.16 cm, 0.23±0.19 cm, 0.26±0.19cm respectively. Conclusion Our results indicate that the delineated targets were comparable for Untag, MaxIP and MeanIP CT with respect to the standard ITV from 4DCT single phase method. The use of the MinIP leads to an underestimation of the contoured volume. The spatial accuracy of the tumor volume is limited to a range within 3 mm (mean distance of the volume centroids) and this leads to a good spatial agreement between PTVs, if generated by expanding an uniformly isotropic 5 mm margin from ITV and GTVs. Among various techniques used for image segmentation, Untag, MaxIP and MeanIP GTVs could be considered as a geometrical surrogate of the standard ITV.