Application of multiobjective optimization concepts in inverse radiotherapy planning

Intensity Modulated Radiotherapy (IMRT) belongs to the most advanced techniques in cancer treatment. Due to the large number of degrees of freedom the inverse problem of IMRT is formulated as a scalar optimization problem. The conventional scalarization approach of the objective function represents an a priori trade-off between the planning goals and can lead to time consuming planning process. New multiobjective approaches try to overcome the difficulties by a precalculation of a set of Pareto optimal treatment plans from which the planner can subsequently choose the best suited plan. In this work we evaluate a new multiobjective treatment planning system. We investigate three different generalized equivalent uniform dose (gEUD) based modeling approaches and study the sensitivities of the corresponding model parameters. Quality measures for entire Pareto optimal databases are developed and applied to clincal cases. In retrospective planning studies we show that the new system is compatible with a clinical reference treatment planning system and that the total planning time for prostate cases can be significantly reduced. In the last part we develop a method to detect the imperative trade-offs in a plan database and apply techniques from linear as well as non-linear dimensionality reduction. They allow meaningful visualizations of high dimensional Pareto fronts and an insight into the underlying trade-offs between the planning goals