Radiation therapy beam angle optimization methods: A comparison study

Radiation therapy for cancer treatment involves selecting appropriate beam angles with the right amount of radiation dose to the tumor cells while sparing the normal tissues surrounding it. Radiation is delivered to the tumor region from a set of discrete angles. Currently, the physicians decide the treatment angles based on their experience. Several optimization techniques have been reported for automating the selection of treatment beam angles. Therefore, the primary goal of this paper is to understand strengths and weaknesses of published optimization methods for the beam selection problem. A collection of six optimization techniques from the literature has been selected, implemented, and compared for computation performance and solution quality. The methods examined are Mixed Integer Programming (MIP), Nested Partitions (NP), Simulated Annealing (SA), Branch and Prune (BP), Genetic Algorithm (GA), and Local Neighborhood Search (LNS). The methods are explained, the evaluation procedure is specified, and then results are compared.