A memetic algorithm for dosimetric optimization in CyberKnife robotic radiosurgical treatment planning

Abstract

A memetic algorithm is proposed for optimizing beam weights for robotic radiosurgical treatments delivered via the CyberKnife (Accuray, Inc., Sunnyvale, CA) system. The fitness function includes terms representing the tumor as well as terms representing organs-at-risk, and is of a quadratic form. Optimization involves inverse treatment planning, during which a set of beam weights is sought such that the user-specified radiation dose distribution is produced by the optimized ensemble of beam weights; the dose to the tumor is maximized, while the doses to the critical structures are minimized. In the present study, four distinct CT data sets for patients with carcinoma of the prostate were used to generate eight treatment plans, so that for each data set, a hypofractionated treatment plan (5 fractions) was created, as well as a treatment plan for a standard protracted dose fractionation schedule (38 fractions). In all cases, the memetic algorithm produced a treatment plan satisfying all clinical criteria in optimization times of 22-46 minutes.