Despite promising developments, patient outcome modeling is challenged by small datasets and geographical biases. Federated learning is a field designed to address these challenges by permitting machine learning models to move between different institutions, generalizing to larger and more diverse data sets, with the ultimate goal of developing more robust models.

A proof-of-concept study will be undertaken using open-source head and neck data to predict patient outcomes using clinical and radiomic data. Clinicians at PM will be engaged to help direct specific clinical research questions.