Multiobjective optimization in radiosurgery : How to approximate and navigate on the Pareto surface

Abstract: Cancer is a common cause of death worldwide and radiotherapy is one of the treatments used. Since treatment planning is a time consuming matter for the radiation therapist, a way to decrease the time spent finding the plan would be an improvement. This can be achieved by precalculating a number of optimal plans and then choosing among these in real-time. In this thesis a dual algorithm for approximation of the Pareto optimal plans suggested by Bokrantz and Forsgren, was adapted to the parameters of the Leksell Gamma Knife®. A Graphical User Interface was also created, based on the navigation tool described by Monz et al to enable choosing among the pre-calculated dose plans. The computational time of the algorithm was investigated and the dimensionality of the solutions and Pareto optimal points were looked at to see if it might be possible to reduce the number of dimensions to speed up computations. Although no certain conclusions can be drawn about dimensionality reduction, I found no reason to rule that possibility out. It was also confirmed that there is reason to keep the number of objectives low to get a better approximation.