ESTRO 38 Abstract book

S491 ESTRO 38

and German Cancer Research Center DKFZ, Heidelberg, Germany Purpose or Objective To investigate the potential of Particle Swarm Optimization (PSO) for automatic planning of VMAT radiotherapy (RT). The aim of this PSO is to formulate an optimal patient individual treatment planning problem, i.e. to define optimal planning constraints leading to the PSO is a statistical, collective and iterative optimization technique which uses the knowledge of each particle about its own best and the swarm’s best position to update particle positions for each generation. For automatic RT planning, a particle is considered a plan with its position determined as a vector of planning constraints. To evaluate plan quality, i.e. determine the particle’s best positions a scoring function was introduced, based on the sum of individual plan quality scores for dose-volume- histogram (DVH) parameters for planning target volume (PTV) and priority organs at risk (OAR). The plan quality score is defined in a way to increase if the DVH parameter fulfills planning goals and penalizes if the goals are violated. Automatic treatment planning using PSO was tested for 8 postoperative prostate cases, with a prescribed dose of 66 Gy to the PTV in addition to two rectum and one bladder constraint. N=30 particles were initialized and the PSO was executed for m=100 generations. PSO and manually generated plans were compared dosimetrically with respect to the planning goals, visual inspection of dose distributions and DVHs. Results PSO successfully proposed treatment plans comparable to manually optimized ones in all 8 cases. The mean (range) PTV EUD was 65.4 Gy (64.7-65.9) for manual and 65.2 Gy (64.8-65.6) for PSO plans, respectively. Also D 98% for the PTV is slightly higher in manual plans, with 62.8 Gy (61.7- 63.6) and 62.5 Gy (61.8-63.0), respectively. On the other hand, PSO plans achieve lower doses in rectum D 2% , 67.0 Gy (66.6-67.5) vs. 66.2 Gy (65.8-66.5, p=0.008). Fig 1 provides an overview of all evaluated planning goals. On average, PSO proposed plans with better OAR sparing but inferior PTV doses. However, this compromise between PTV and OAR doses is directly related to the definition of the scoring function. According to the scoring function used in this study, manual plans had lower plan quality scores compared to PSO plans with -0.61 (-2.41-0.96) vs. 1.41 (-0.84-6.78). Overall, good convergence of the planning goals and the plan quality score was observed with increasing particle generations (Fig 2a-b). As demonstrated by Fig 2c-d, PSO proposed comparable dose distributions. best achievable plan. Material and Methods

Conclusion A virtual bolus of 1.0 cm and a HU-value of approximately -500 HU would maximize plan robustness against movements up to 1.0 cm in the breathing direction. Additionally, this HU-value minimizes the dosimetric impact of the strategy. Therefore in most of the patients it will be no necessary to re-normalize the clinical treatment and if it was, it will be a minor change. The study has been focus at breast radiotherapy but same methodology could be applied on other locations. PO-0919 Automatic radiotherapy treatment planning using Particle Swarm Optimization L.A. Künzel 1 , S. Leibfarth 1 , O.S. Dohm 2 , A. Müller 2 , D. Zips 2,3 , D. Thorwarth 1,3 1 Section for Biomedical Physics Department for Radiation Oncology, University Hospital Tübingen, Tübingen, Germany ; 2 Department for Radiation Oncology, University Hospital Tübingen, Tübingen, Germany ; 3 German Cancer Consortium DKTK partner site Tübingen,