Additive MIL: Intrinsically Interpretable Multiple Instance Learning for Pathology

TL;DR

This paper introduces Additive MIL, a transparent model for pathology that allows exact spatial credit assignment, improving interpretability and trust in clinical applications without sacrificing predictive accuracy.

Contribution

The authors propose a simple modification to existing MIL models to make them additive and interpretable, aligning model explanations with pathologists' diagnostic regions.

Findings

Additive MIL enables exact spatial credit assignment.

The model's explanations match pathologists' regions of interest.

It improves interpretability over classical attention heatmaps.

Abstract

Multiple Instance Learning (MIL) has been widely applied in pathology towards solving critical problems such as automating cancer diagnosis and grading, predicting patient prognosis, and therapy response. Deploying these models in a clinical setting requires careful inspection of these black boxes during development and deployment to identify failures and maintain physician trust. In this work, we propose a simple formulation of MIL models, which enables interpretability while maintaining similar predictive performance. Our Additive MIL models enable spatial credit assignment such that the contribution of each region in the image can be exactly computed and visualized. We show that our spatial credit assignment coincides with regions used by pathologists during diagnosis and improves upon classical attention heatmaps from attention MIL models. We show that any existing MIL model can be made additive with a simple change in function composition. We also show how these models can debug model failures, identify spurious features, and highlight class-wise regions of interest, enabling their use in high-stakes environments such as clinical decision-making.