[Impact of DVH Outliers Registered in Knowledge-based Planning on Volumetric Modulated Arc Therapy Treatment Planning for Prostate Cancer].

ABSTRACT

RapidPlan, a knowledge-based planning software, uses a model library containing the dose-volume histogram (DVH) of previous treatment plans, and it automatically provides optimization objectives based on a trained model to future patients for volumetric modulated arc therapy treatment planning. However, it is unknown how DVH outliers registered in models influence the resulting plans. The purpose of this study was to investigate the effect of DVH outliers on the resulting quality of RapidPlan knowledge-based plans generated for patients with prostate cancer. First, 123 plans for patients with prostate cancer were used to populate the initial model (modelall). Next, modelall-20 and modelall-40 were created by excluding DVH outliers of bladder optimization contours 20 and 40 patients from modelall, respectively. These models were used to create plans for a 20-patient. The plans created using modelall-40 showed reductions of D30% and D50% in the bladder wall dose, and the DVH shape excluding outliers were affected. However, there were no significant differences in monitor units, target doses, or bladder wall doses between each treatment plan. Thus, we have shown that removal of DVH outliers from models does not affect the quality of plans created by the model.