ESTRO 2022 - Abstract Book

S576

Abstract book

ESTRO 2022

Conclusion For rectal cancer VMAT, a machine learning automated planning model can be shared across institutes while conforming to the local clinical goals without laborious retraining, but by applying a generic postprocessing strategy. The resulting MLO plans for rectal cancer VMAT were non-inferior to the manual plans for the majority of patients (79%).

MO-0642 Comparison of different approaches for automatic plan adaptation in MR-guided radiotherapy

B. Tengler 1 , M. Hagmüller 1 , L.A. Künzel 2 , D. Zips 3 , D. Thorwarth 1

1 University Hospital and Medical Faculty. Eberhard Karls University Tübingen, Department of Radiation Oncology, Section for Biomedical Physics, Tübingen, Germany; 2 University Hospital and Medical Faculty. Eberhard Karls University Tübingen, Department of Radiation Oncology, Section for Biomedical Physics , Tübingen, Germany; 3 University Hospital and Medical Faculty. Eberhard Karls University Tübingen, Department of Radiation Oncology, Tübingen, Germany Purpose or Objective MR-guided radiotherapy (MRgRT) provides the opportunity of daily plan adaptation based on information about the patient’s current anatomy. Automatic MRgRT planning approaches may be used to realize fast plan optimization during online adaptation. The aim of this work was to evaluate the differences in terms of dosimetric quality comparing automatic online plan adaptation starting from a reference plan to complete re-optimization both based on the anatomy of the day. Materials and Methods Clinical data of ten prostate carcinoma patients treated with online adaptive MRgRT (20 x 3 Gy) at the 1.5 T MR-Linac were included into this study. Optimization of the MRgRT plans was carried out by an automated planning approach using a particle swarm optimization (PSO). For each treatment fraction, two adapted plans were optimized using (1) a plan re- optimization with the optimal planning constraints achieved for the reference PSO plan at baseline and (2) a completely new automatic PSO planning approach to take the anatomical variations best possible into account. Method 1 consists of a fast plan optimization exploring a limited search space defined by a set of plan constraints determined upfront whereas method 2 allows for unlimited exploitation of plan constraints for optimal adaptation at the cost of increased calculation times. The daily dose distributions obtained with the two automatic planning approaches were evaluated by comparing the relative adherence to a set of clinically relevant dosimetric criteria (cf. table 1). The significance of the differences was considered using a t-test at the 5% significance level.