A Feasible Solution to the Beam-Angle-Optimization Problem in Radiotherapy Planning With a DNA-Based Genetic Algorithm

Abstract

Intensity-modulated radiotherapy (IMRT) is now becoming a powerful clinical technique to improve the therapeutic radio for cancer treatment. It has been demonstrated that selection of suitable beam angles is quite valuable for most of the treatment plans, especially for the complicated tumor cases and when limited number of beams is used. However, beam-angle optimization (BAO) remains a challenging inverse problem mainly due to the huge computation time. This paper introduced a DNA genetic algorithm (DNA-GA) to solve the BAO problem aiming to improve the optimization efficiency. A feasible mapping was constructed between the universal DNA-GA algorithm and the specified engineering problem of BAO. Specifically, a triplet code was used to represent a beam angle, and the angles of several beams in a plan composed a DNA individual. A bit-mutation strategy was designed to set different segments in DNA individuals with different mutation probabilities; and also, the dynamic probability of structure mutation operations was designed to further improve the evolutionary process. The results on simulated and clinical cases showed that DNA-GA is feasible and effective for the BAO problem in IMRT planning, and to some extent, is faster to obtain the optimized results than GA.