Machine Learning for Modeling Dose-Related Organ-at-Risk Complications after Radiation Therapy

Purpose: To predict organ-at-risk (OAR) complications as a function of dose-volume (DV) constraint settings without explicit plan computation in a multi-plan IMRT framework. Methods and Materials: A large number of plans were generated by varying the DV constraints (input features) on the OARs (multi-plan framework), and the OAR complications in the plans (plan properties) were modeled as a function of the imposed DV constraint settings, which were used as input to machine learning (ML) algorithms. These ML approaches were used to model two OAR complications following head-and-neck and whole pelvis/prostate intensity-modulated radiation therapy, xerostomia and grade 2 rectal bleeding. Two-fold cross-validation was used for model verification and mean errors were reported. Results: In the head and neck case, the mean absolute prediction error of the saliva flow rate normalized to the pre-treatment saliva flow rate was 0.42% with a 95% confidence interval of [0.41%, 0.43%]. In the whole pelvis/prostate case, an average prediction accuracy of 97.04% with a 95% confidence interval of [96.67%, 97.41%] was achieved for grade 2 rectal bleeding complications. Conclusion: ML can be used for predicting OAR complications during treatment planning allowing for alternative DV constraint settings to be assessed within the planning framework.