MSDS Alumnus Presents Machine Learning Research on Gamma Knife Radiosurgery at International Conference

Brett Castro, a 2025 M.S. in Data Science graduate at the University of Virginia, represented his capstone team in November at the 2025 World Federation of Neuro-Oncology Societies Scientific Meeting in Honolulu, where he shared findings that apply machine learning to improve understanding of Gamma Knife radiosurgery dose and patient outcomes.

The project was developed in collaboration with the UVA Health Gamma Knife clinical team and Associate Professor of Data Science Adam Tashman, who served as faculty mentor. Using a large, multi-decade clinical dataset, the team explored how different modeling approaches could predict radiation dose and analyze treatment outcomes. Tree-based machine learning models, including XGBoost, consistently outperformed linear baselines in dose prediction tasks, while analyses of tumor progression and follow-up treatment highlighted persistent challenges related to class imbalance and the need for expanded datasets and future validation.

Presenting the work at WFNOS SNO 2025 offered Castro an opportunity to engage directly with clinicians and researchers in neuro-oncology, situating their findings within real-world clinical workflows. Feedback and discussion at the conference helped contextualize both the promise and limitations of machine learning approaches in high-stakes medical decision-making.

The capstone team included Isaac Levy, Tyler Gorecki, and Muhurto Rahman, with support from collaborators across UVA Health and the UVA School of Data Science. Their work reflects the School’s emphasis on applied, interdisciplinary research that connects data science methods with meaningful societal and clinical impact.

Castro is a second lieutenant in the United States Space Force undergoing officer space training.