303-TITAN: Multimodal Whole-Slide Pathology Foundation Model

The paper describes the development and extensive evaluation of the TITAN (Transformer-based pathology Image and Text Alignment Network), a new multimodal foundation model for computational pathology. TITAN is pretrained on a massive dataset of over 335,000 whole-slide images (WSIs) using a combination of vision-only learning and subsequent vision-language alignment with both synthetic fine-graine...去小宇宙查看完整单集简介
前往小宇宙评论区与主播互动