Evolutionary Large‐Scale Multi‐Objective Optimization in Radiotherapy Planning
Xingyi Zhang, Ran Cheng, Ye Tian, Yaochu Jin · 2024
This chapter introduces an application of evolutionary large-scale multi-objective optimization, namely, intensity-modulated radiotherapy (IMRT) planning. Mathematical programming methods, heuristics, and deep learning methods are incapable of handling the conflicting objectives for providing diverse treatment plans, while general evolutionary algorithms are ineffective to optimize the large number of variables within a limited number of function evaluations. Therefore, several multi-objective evolutionary algorithms (MOEAs) have been developed for IMRT planning with customized search strategies, such as the bi-encoding scheme and local search strategies of multi-objective evolutionary algorithm/IMRT. Large-scale MOEAs are more effective than conventional optimizers including general MOEAs for IMRT planning. However, the research of evolutionary multi-objective optimization focuses on continuous optimization, while MOEA/IMRT can improve the performance on IMRT planning by developing search strategies for combinatorial optimization.