Project title: Uncertainty-Driven Triage Between Cheap and Expensive Imaging

Brief project description: This project investigates how artificial intelligence can support clinical decisions about when additional, costly medical tests are necessary. Rather than predicting a diagnosis from an accessible imaging modality, the research aims to determine whether an expensive test, such as MRI, CT, or PET, would provide information that changes patient management. The first phase develops a calibrated model to estimate the likelihood of such changes while accounting for test costs and clinical context. The second phase explores generative models to simulate plausible test outcomes and assess their impact on treatment decisions. The goal is to improve diagnostic efficiency, resource allocation, and patient care.