＊AI for digital pathology image analysis provides the pathway to achieve personal precision medicine

＊Pathology image analysis encounters the problems of high labeling cost and image properties drift in different hospitals

＊This project aims to build a FL-based collaborative training platform embedded with techniques to resolve the aforementioned problems

1. Robust Deep Learning

＊Learning against data distribution shift

＊Optimized learning based on inaccurate data

＊Explainable AI model

2. Self-learning image detection

＊Continuous learning for object detection

＊Continuous learning based on self-annotation data

3. Less-annotation learning

＊Semi-supervised learning

＊ Weakly supervised learning

4. Open AI Pathology Platform

＊Data dynamic agreement

＊Data governance and model sharing mechanism