Built during a Microsoft hackathon track: deep learning for pattern recognition, classification, and instance segmentation on uterine cancer pathology slides — aimed at helping pathologists spot and characterize abnormal tissue faster.
Why it matters
Pathology workloads are high-stakes and time-intensive. Models that assist with localization and classification do not replace clinical judgment; they shorten the path from slide to a reviewed finding. This project focused on demonstrating that assistive loop in a competition setting.