A Rapid Learning Approach to the Knowledge Modeling of Radiation Therapy Planning

Abstract

The purpose of this study is to implement a rapid learning method to train the knowledge models to predict the organ-at-risk (OAR) dose sparing in radiation therapy (RT) based on an array of patient anatomical features. We also aim to establish the evaluation criteria and validation method to ensure an accurate and efficient learning process. A rapid learning approach is utilized to train the knowledge models in this study. 100 clinical cancer cases in the pelvic region were retrospectively analyzed. Among them, 40 cases are low-tointermediate risk prostate cases (Type I), 20 are high-risk prostate cases with lymph node irradiation (Type II), 40 are anorectal cancer cases (Type III). Starting from a base model for type I cases, increasing number of cases with more complex planning-target-volume (PTV)-OAR anatomies (type II and type III) were continuously added into the training case pool. The studentized residual and the leverage values are calculated as evaluation criteria at each step. The efficiency and accuracy of the learning method was quantified by the learning curve. The gEUD in the bladder and rectum are compared between the model predictions and actual values for the validation cases. The Median of the Absolute value of their Differences (MAD) are calculated for the validation cases. The MAD of the predicted OAR gEUD in all three types of cases gradually decreases when increasing number of training cases are added in training. The knowledge models learned by this method reach comparable level of prediction accuracies as in batch training mode even less training cases. The rapid learning approach is able to learn knowledge models for multiple cancer types in the pelvic region with comparable accuracy to the batch training method and with improved efficiency. This approach will facilitate the implementation of the knowledge based radiation therapy planning in clinics.