Anatomy-based three-dimensional dose optimization in brachytherapy using multiobjective genetic algorithms

In conventional dose optimization algorithms, in brachytherapy, multiple ob jectives are expressed in terms of an aggregating function which combines i ndividual objective values into a single utility value, making the problem single objective, prior to optimization. A multiobjective genetic algorithm (MOGA) was developed for dose optimization based on an a posteriori approa ch, leaving the decision-making process to a planner and offering a represe ntative trade-off surface of the various objectives. The MOGA provides a fl exible search engine which provides the maximum of information for a decisi on maker. Tests performed with various treatment plans in brachytherapy hav e shown that MOGA gives solutions which are superior to those of traditiona l dose optimization algorithms. Objectives were proposed in terms of the CO W distribution and differential volume histograms, taking into account pati ent anatomy in the optimization process. (C) 1999 American Association of P hysicists in Medicine. [S0094-2405(99)00309-0].