AGE-MIL: Anchor-Guided Evidence Learning for Patient-Level Prediction

2026-06-10

AI Summary

Current computational pathology approaches predominantly rely on multiple-instance learning (MIL) applied to individual whole-slide images (WSIs), which struggles to effectively model patient-level prediction tasks requiring integration across multiple WSIs. This limitation often leads to unstable optimization and insufficient reliability under weak supervision. To address this, this work proposes the AGE-MIL framework, which introduces a patient-level anchoring mechanism to guide the retrieval and fusion of diagnostically relevant local regions. It further formalizes patient risk prediction as an evidence accumulation process, aligning clinical diagnostic reasoning with model optimization objectives. By integrating anchor-guided attention, evidential deep learning, and weakly supervised training, AGE-MIL significantly outperforms eight state-of-the-art MIL methods across six patient-level prediction tasks in two independent cohorts, achieving both superior performance and enhanced stability.

Abstract

Existing computational pathology methods predominantly operate within whole-slide image (WSI)-level multiple instance learning (MIL) paradigms, while patient-level modeling remains underexplored. In routine pathological practice, however, pathologists derive diagnostic and prognostic conclusions by integrating evidence across multiple WSIs rather than relying on any single slide. This discrepancy creates a fundamental misalignment when patient-level supervision is directly imposed on conventional MIL frameworks, often leading to unstable optimization and degraded predictive reliability. To address this issue, we propose Anchor-Guided Evidence MIL (AGE-MIL), a weakly supervised framework for patient-level prediction. AGE-MIL constructs a patient-level anchor from slide representations to capture global pathological context and guide the retrieval and integration of diagnostically relevant local patches, enabling robust patient-level modeling. Patient-level risk is further modeled as an evidence accumulation process, promoting stable optimization under weak supervision. AGE-MIL is evaluated on six clinically relevant patient-level prediction tasks from two independent cohorts. Experimental results show that the proposed framework consistently outperforms eight state-of-the-art MIL methods. Code is available at https://github.com/wodeniua/AGE-MIL.

Problem

• patient-level prediction
• multiple instance learning
• computational pathology
• weak supervision
• whole-slide image

Innovation

• Anchor-Guided
• Evidence Learning
• Patient-Level Prediction
• Multiple Instance Learning
• Computational Pathology