Multicriteria Optimization in Inverse Radiotherapy Planning

The development of intensity-modulated radiotherapy (IMRT) has greatly enhanced the possibilities of shaping the dose distribution. However, because of physical limitations, even with IMRT a certain dose to organs at risk is unavoidable, so in treatment planning often a compromise between different plan criteria has to be found. This thesis presents a new optimization concept that is dedicated to this type of problem. It is based on the fact that there is no single optimal plan, but many so-called Pareto optimal plans, which cannot be improved in one criterion without worsening another. A database with all Pareto optimal plans is generated and the decision which compromise is clinically the best is left to the planner. Because of the higher calculation complexity, the dose calculation of the existing planning system was accelerated by a factor of three through statistical sampling of the calculation matrices. To describe a plan, a vector containing the 'equivalent uniform doses' (EUDs) of each structure is used. Therefore two EUD models were analyzed, the 'max & mean' and the 'generalized mean' model. An algorithm was developed to include the 'generalized mean' model in the plan optimization. At the end of the thesis, the database containing all Pareto optimal solutions of a clinical test case with 1 target volume and 2 organs at risk was generated and visualized. The planner can see at first sight the dependencies of the different plan elements and can decide which plan is clinically the best