Automation of Contouring and Planning in Radiotherapy

This thesis focuses on automated contouring of planning CT-scans and automated treatment plan generation in radiotherapy to improve treatment quality and / or efficiency compared to current manual procedures. For automated contouring we have evaluated a commercially available atlas-based autosegmentation tool. With this tool, the CT-scan of a new patient can be segmented based on segmented scans of previously treated patients, called atlases, thereby significantly reducing the manual contouring time. For automated plan generation we have investigated Erasmus-iCycle, an in-house developed algorithm for multicriterial beam angle and profile optimization. For individual patients, Erasmus-iCycle can fully automatically generate Pareto-optimal plans, based on a wish-list with constraints and prioritized objectives specified for the corresponding disease / tumor site. Compared to manually generated plans, Erasmus-iCycle plans are superior or equal in quality. Moreover, cost savings are possible, as time consuming manual planning by dosimetrists can be avoided. Feasibility of high quality automated contouring and planning has been demonstrated, which is also highly relevant for the rapidly evolving field of adaptive daily re-planning.