Unsupervised Learning Approach for Detailed Cell Type Mapping in H&EStained Images

Time: 4:00 pm
day: Day Two


  • Generating datasets that accurately and exhaustively delineate cells in H&E images is challenging and time-consuming, particularly for achieving comprehensive coverage of the entire cellulome
  • We propose a machine learning framework that utilizes state-of-the-art cell segmentation models in conjunction with a modified self-supervised learning model to capture intricate embeddings of individual cells in H&E whole slide images
  • Our framework is designed to cluster similar cell types, enabling their differentiation in an automated manner. This paves the way for comprehensive, automated classification of cell types in H&E-stained images