Implementation of gradient projection algorithm to radiation therapy treatment planning

The fluence map optimization (FMO) is a key element in radiation therapy treatment planning. In this study, we implement two types of gradient projection algorithm to a dose-based objective function formulated as a bound-constrained quadratic programming (BCQP) problem. The purpose of this implementation is to assess the validity and convergence properties of these algorithms. In addition, we investigate the possibility of being trapped in a local minimum when using different initial intensity distributions. We use two cancer cases to illustrate the performance of these algorithms, including one prostate and one head-and-neck case. Our preliminary results indicate that the algorithms are producing clinically acceptable plans in a reasonable amount of time, and essentially converge to the same plans regardless of the starting point used in the algorithms.