Compressed Sensing-Based LDR Brachytherapy Inverse Treatment Planning with Biological Models

Abstract

New compressed sensing-based planning algorithms allow for fast computations of optimal planning results in low-dose-rate (LDR) brachytherapy. This enables to integrate complex models in the planning process. In this paper, we develop a new strategy for including a biological model on tumor control probability (TCP) and normal tissue complication probability (NTCP) into the objective function for plan optimization. These models were tested on clinical prostate cancer cases for their effects on the planning results relative to standard physical dose constraints for planning as reference. Interestingly, with weighting treatment risks, we observe plans using biological models assign more dose to the urethra since it is less radiation sensitive than the rectum whereby the latter is spared in order to reduce side effects. At the same time, the overall TCP is comparable. We conclude that the standard plan quality evaluation based on physical dose alone does not easily allow correctly assessing treatment risks. Hence, biological models for LDR brachytherapy treatment planning are a promising approach for an optimal management of treatment outcomes of brachytherapy.