SMC Digital Pathology Consortium

Development of integrated analysis platform based on digital pathology data of high incidence cancer in Korea

This project aims to build an AI-based cancer analysis platform using digital pathology data. Beyond analysis techniques limited to certain brain tumors with restricted applicability, we first train a model for image classification of neuroendocrine tumors based on grade. Additionally, we develop technologies to utilize explainable AI and visualize the rationale behind the results of deep learning models, making it applicable to pathological diagnosis. Subsequently, we develop a tool to assist in quality validation of pathological images by training a deep learning model for image segmentation, visualizing areas with defects in tissue slide images, thus reducing the burden of manually excluding low-quality pathological images. Finally, we propose a methodology to train and obtain deep learning classification models stably from high resolution pathological image data..