Contrastive Integrated Gradients: A Feature Attribution-Based Method for Explaining Whole Slide Image Classification

TL;DR

This paper introduces Contrastive Integrated Gradients (CIG), a new feature attribution method for Whole Slide Image classification that improves interpretability by highlighting class-discriminative regions and satisfying theoretical axioms.

Contribution

CIG enhances interpretability of WSI models by computing contrastive gradients, providing sharper class distinctions and introducing new metrics for attribution quality evaluation.

Findings

CIG produces more informative and visually aligned attributions.

CIG outperforms baseline methods on multiple cancer datasets.

The proposed metrics MIL-AIC and MIL-SIC effectively measure attribution quality.

Abstract

Interpretability is essential in Whole Slide Image (WSI) analysis for computational pathology, where understanding model predictions helps build trust in AI-assisted diagnostics. While Integrated Gradients (IG) and related attribution methods have shown promise, applying them directly to WSIs introduces challenges due to their high-resolution nature. These methods capture model decision patterns but may overlook class-discriminative signals that are crucial for distinguishing between tumor subtypes. In this work, we introduce Contrastive Integrated Gradients (CIG), a novel attribution method that enhances interpretability by computing contrastive gradients in logit space. First, CIG highlights class-discriminative regions by comparing feature importance relative to a reference class, offering sharper differentiation between tumor and non-tumor areas. Second, CIG satisfies the axioms of integrated attribution, ensuring consistency and theoretical soundness. Third, we propose two attribution quality metrics, MIL-AIC and MIL-SIC, which measure how predictive information and model confidence evolve with access to salient regions, particularly under weak supervision. We validate CIG across three datasets spanning distinct cancer types: CAMELYON16 (breast cancer metastasis in lymph nodes), TCGA-RCC (renal cell carcinoma), and TCGA-Lung (lung cancer). Experimental results demonstrate that CIG yields more informative attributions both quantitatively, using MIL-AIC and MIL-SIC, and qualitatively, through visualizations that align closely with ground truth tumor regions, underscoring its potential for interpretable and trustworthy WSI-based diagnostics