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

2025-03-02

AI Summary

Accurate subtyping of malignant lymphoma is critical for treatment planning and prognostic assessment, yet existing methods struggle to balance classification accuracy with clinical interpretability. To address this, we propose an interpretable multi-instance learning framework that—uniquely—integrates cellular graph structures with whole-slide images (WSIs) in a dual-modal representation. Our method employs a customized Mixture-of-Experts (MoE) mechanism to automatically localize subtype-specific regions and explicitly model the spatial distribution of key cell types. Evaluated on a large-scale dataset of 1,233 WSIs, it achieves state-of-the-art classification accuracy, significantly outperforming ten baseline methods. Moreover, it delivers both region-level and cell-level visual explanations aligned with pathologists’ domain knowledge, ensuring clinical verifiability and translational relevance.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Interpretability, Explainability, and Transparency
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Malignant lymphoma subtype classification directly impacts treatment strategies and patient outcomes, necessitating classification models that achieve both high accuracy and sufficient explainability. This study proposes a novel explainable Multi-Instance Learning (MIL) framework that identifies subtype-specific Regions of Interest (ROIs) from Whole Slide Images (WSIs) while integrating cell distribution characteristics and image information. Our framework simultaneously addresses three objectives: (1) indicating appropriate ROIs for each subtype, (2) explaining the frequency and spatial distribution of characteristic cell types, and (3) achieving high-accuracy subtyping by leveraging both image and cell-distribution modalities. The proposed method fuses cell graph and image features extracted from each patch in the WSI using a Mixture-of-Experts (MoE) approach and classifies subtypes within an MIL framework. Experiments on a dataset of 1,233 WSIs demonstrate that our approach achieves state-of-the-art accuracy among ten comparative methods and provides region-level and cell-level explanations that align with a pathologist's perspectives.

Problem

• Develops explainable classifier for lymphoma subtype classification.
• Integrates cell graph and image features for accurate subtyping.
• Provides region-level and cell-level explanations for pathologists.

Innovation

• Multi-Instance Learning for lymphoma subtyping
• Cell graph and image feature fusion
• Explainable ROI identification via Mixture-of-Experts