SURG-88. Optimal radiosurgical dose planning for tumor control: A machine learning approach to predicting Gamma Knife treatment and outcomes

Abstract Gamma Knife Radiosurgery offers a non-invasive treatment option for treating brain metastases, which are often in surgically inoperable regions. However, determining optimal dosing regimens remains a clinical challenge due to patient outcome variability. In this study, we propose a tree-based machine-learning framework for personalized dose recommendations, tumor prediction progression, and follow-up treatment assessment using a dataset of over a thousand metastases treated at the UVA Health Gamma Knife Center between 1980 and 2025. We incorporate heterogeneous patient data from three distinct cancer types (breast, melanoma, and renal cell carcinoma), including histology, genetic markers, tumor location, and volume, to feature engineer variables into three predictive models. Our best-performing dose classification model achieved F-1 scores as high as 0.82 for breast cancer, while regression demonstrated low RMSE values at around 1.5 Gy. Models for tumor progression and patient-followup likelihood were impacted by class imbalance, however, overall, these models underscore the viability of data-driven approaches to support individualized Gamma Knife radiosurgical planning with positive impacts for improving clinical decision-making and patient outcomes.