ESTRO 2020 Abstract Book

S109 ESTRO 2020

methods were compared for the optimization of a lung tumor proton therapy treatment (PTV volume of 195 cm³). The treatment plan is composed of 3 beams and 9064 spots. Results Optimization results are summarized in Figure 1 and Table 1 . For both methods, the final dose was calculated with MCsquare by simulating 34 millions particles in order to reduce the MC statistical noise to less than 2%. Both methods achieved similar target coverage and organ sparing. However, the beamlet-free algorithm offered better performances by significantly reducing the computation time (reduced by 59%) and RAM memory requirements (reduced by 91%).

Conclusion Our study demonstrates that LSTM networks are suited for fast and accurate dose calculation of proton pencil beams within highly heterogeneous geometries. This provides stark motivation to further study the generalization of our results to other proton energies and patient cases. OC-0216 Efficient optimization of Monte Carlo based proton therapy plans using a beamlet-free algorithm K. Souris 1 , A. Barragan 1 , G. Buti 1 , M. Cohilis 1 , G. Janssens 2 , E. Sterpin 1 , J.A. Lee 1 1 UCLouvain, Miro, Brussels, Belgium ; 2 IBA s.a., Research, Louvain-la-Neuve, Belgium Purpose or Objective The optimization of proton therapy treatment plans is often costly in terms of computation time and memory requirements. This is due to the necessity to compute a large number of beamlets (3D dose distribution for each spot). The computation of such beamlets is especially time consuming when Monte Carlo (MC) dose calculation is needed. Moreover, this issue becomes very impractical for robust optimization due to the number of beamlets that grows linearly with the number of uncertainty scenarios considered. In this study, we designed a beamlet-free algorithm for the optimization of proton therapy plans that reduces the memory usage and computation time. Material and Methods Beamlet-based optimization algorithms typically comprise two steps. Beamlets are first pre-computed, preferably using a MC dose engine. Afterwards, spot weights are iteratively adjusted until the weighted sum of all beamlets results in an optimal dose distribution (i.e. the objective function that encodes clinical objectives is minimized). In contrast, our beamlet-free algorithm optimizes the treatment plan in one step and involves only a single Monte Carlo simulation. After the simulation of each particle, the algorithm evaluates if it contributed positively or negatively to the objective function and adjusts the corresponding spot weight accordingly. Thereby, the objective function is evaluated with each new particle, but only considering the traversed voxels at each time. This micro-optimization approach is similar to the stochastic gradient descent used in machine learning. The beamlet-free algorithm was implemented in the open source MC code MCsquare. The beamlet-based optimization was performed by the open source treatment plan optimization application MIROpt, which also uses MCsquare for the MC calculation of beamlets. Both

Conclusion The beamlet-free algorithm proposed for treatment plan optimization reduced the computation time and memory usage. These increased performances can benefit to online plan optimization for adaptive therapy. It could also enable the optimization of more comprehensive robustness solutions and more complex modalities such as proton arc therapy. This new optimization approach is not limited to proton therapy and could be applied to any MC dose calculation. OC-0217 Commissioning of IDEAL/GATE-RTion for Proton and Carbon ion Independent Dose Calculation (IDC) L. Grevillot 1 , D. Boersma 2 , R. Gonzalo Gleyzes 1 , L. Scheuchenpflug 3 , A. Carlino 1 , A. Elia 1 , H. Fuchs 4 , M. Stock 1 1 EBG MedAustron GmbH, Medical Physics, Wiener Neustadt, Austria ; 2 ACMIT GmbH, Acmit, Wiener Neustadt, Austria ; 3 University of Vienna, Faculty of Physics- Isotope Physics, Vienna, Austria ; 4 Medical University of Vienna, Department of Radiation Oncology & Christian Doppler Laboratory, Vienna, Austria Purpose or Objective GATE-RTion is a clinical GATE/GEANT4 release since May 2018. The Independent DosE cAlculation for Light ion beam therapy (IDEAL) project includes GATE-RTion as dose engine. It focuses on the implementation of a CE-marked IDC system for facilities equipped with Scanned Ion Beam Delivery systems. A first prototype is available since 2018 and was transferred in 2019 into a clinical environment of a Light Ion Beam Therapy center. This work presents the