ESTRO 2021 Abstract Book

S659

ESTRO 2021

Conclusion The algorithm can detect relevant OAR delineation outliers in about 10% of clinically treated patients. The algorithm is not based on any prior information but utilises the local institution's delineation practice. Furthermore, the algorithm does not use any image information; therefore, it is suited for both CT and MR based treatment planning. The process is feasible for adaptive online treatment since the evaluation of a full set of contours takes less than half a minute. Local clinical implementation of the code, which will be available as open-source, is ongoing PD-0825 Contribution of PET and CT Images to Machine Learning for Contouring GTVs in Head and Neck Cancer A. Theljani 1 , K. A. Vallis 1 , M. Mark J. Gooding 2 1 Oxford Institute for Radiation Oncology, Department of Oncology, Oxford, United Kingdom; 2 Mirada Medical Ltd, Mirada Medical Ltd, Oxford, United Kingdom

Purpose or Objective