Session

Overview: This symposium focusses on quality assurance of autosegmentation models and AI-based auto-planning for photon and proton therapy. How robust are they? Are they still adequate after a software update or a change in imaging device? During routine clinical application, do they behave as expected, and how can the clinical benefit be evaluated? What biases are introduced and does this change over time? The importance of estimating uncertainty of autosegmentations is discussed as well as the limitations of the Dice Similarity Coefficient. Finally, the question is considered if the dosimetric impact of contouring variations can be used to minimize manual corrections.