BI-RADS Rationale Generation for Breast MRI Segmentation
October 22, 2026
A new framework derives clinically relevant descriptors from DCE-MRI segmentation masks using BI-RADS guidelines. These descriptors are fed into an LLM to generate natural-language rationales, improving the explainability of tumor segmentation for radiologists.
HOW THIS AFFECTS YOU
●
researcherYou can use this structured approach to bridge the gap between pixel-level segmentation and clinical logic.