<ul><li>A novel explainable Multi-Instance Learning (MIL) framework is proposed for malignant lymphoma subtype classification.</li><li>The framework integrates cell distribution characteristics and image information to identify subtype-specific Regions of Interest (ROIs).</li><li>The proposed method achieves high-accuracy subtyping by fusing cell graph and image features using a Mixture-of-Experts (MoE) approach.</li><li>Experiments demonstrate that the approach achieves state-of-the-art accuracy and provides region-level and cell-level explanations.</li></ul>

Explainable Classifier for Malignant Lymphoma Subtyping via Cell Graph and Image Fusion

Discover more