Medial Axis Seeding of a Guided Evolutionary Simulated Annealing (GESA) Algorithm for Automated Gamma Knife Radiosurgery Treatment Planning

Abstract

We present a method to optimize Gamma Knife TM (Elekta, Stockholm, Sweden) radiosurgery treatment planning. A Guided Evolutionary Simulated Annealing optimization algorithm is used to maximize the therapeutic benefit through a probability model that dissects a patient volume image into three components: normal, critical normal, and tumor tissue. This evolutionary optimization algorithm may be seeded randomly or via an automatically detected medial axis. We use indices of dose conformality, level, and homogeneity to evaluate the degree to which a treatment plan has been optimized. Two clinical examples compare the GESA algorithm with current manual methods. GESA optimization shows therapeutic advantage over the treatment team's manual effort. We find that computation of treatment plans with more than 8 shots require initial medial axis seeding (i.e., shot: number, size, and position) to complete within 8 hours on our workstation.