Room: ePoster Forums
Purpose: To optimize the selection of beam angels for proton beam therapy based on genetic algorithm(GA) and evaluate the effectiveness of proposed algorithm using clinical cases.
Methods: One hundred eight proton therapy liver plans, which were randomly selected from Proton Therapy Center, Samsung Medical Center in Korea, were used for GA to select suitable beam angles. A chromosome represents the combination of beam angles, and each gene in the chromosome represents a beam angle. A fitness function was defined to include quality assessment based on dose-volume histograms, and a penalty item to constrain the maximum allowable doses delivered to critical organs. Genetic operations such as selection, crossover, mutation, and replacement have been used to evolve the population and were repeated until fitness has reached the minimum criteria. Treatment plans generated by the GA using suggested fitness function were compared with proton dosimetrist(PD) treatment plans in terms of planning time and dosimetry. Qualitative plan quality generated by proposed GA was evaluated by expert proton therapy physicists and oncologists using a questionnaire.
Results: To verify the proposed GA, five new clinical cases were selected and evaluated. The average generation to select suitable beam angles in five clinical cases was less than 21 times which were compatible for the preparation time of actual treatment plan in clinic. The results show that generated plans by GA were dosimetrically equivalent to the PD plans although there was a small difference in DVH. Respondents ranked the GA based plans as equivalent to expert PD treatment plans in terms of various clinical factors.
Conclusion: Our results indicate that proton liver treatment plans that have equivalent quality to plans created by professional dosimetrist can be automatically generated using proposed GA. The details of results for the application of GA in proton therapy planning will be presented.