Agent Aggregator with Mask Denoise Mechanism for Histopathology Whole Slide Image Analysis

TL;DR

This paper introduces AMD-MIL, a novel weakly supervised learning method for histopathology WSI analysis that enhances attention mechanisms with agent tokens and mask denoising to improve classification accuracy and interpretability.

Contribution

The paper proposes AMD-MIL, which incorporates agent tokens and a mask denoise mechanism to better capture inter-instance information and reduce noise in WSI classification.

Findings

AMD-MIL outperforms state-of-the-art methods on multiple datasets.

It effectively captures micro-metastases in cancer detection.

The approach improves interpretability of WSI analysis.

Abstract

Histopathology analysis is the gold standard for medical diagnosis. Accurate classification of whole slide images (WSIs) and region-of-interests (ROIs) localization can assist pathologists in diagnosis. The gigapixel resolution of WSI and the absence of fine-grained annotations make direct classification and analysis challenging. In weakly supervised learning, multiple instance learning (MIL) presents a promising approach for WSI classification. The prevailing strategy is to use attention mechanisms to measure instance importance for classification. However, attention mechanisms fail to capture inter-instance information, and self-attention causes quadratic computational complexity. To address these challenges, we propose AMD-MIL, an agent aggregator with a mask denoise mechanism. The agent token acts as an intermediate variable between the query and key for computing instance importance. Mask and denoising matrices, mapped from agents-aggregated value, dynamically mask low-contribution representations and eliminate noise. AMD-MIL achieves better attention allocation by adjusting feature representations, capturing micro-metastases in cancer, and improving interpretability. Extensive experiments on CAMELYON-16, CAMELYON-17, TCGA-KIDNEY, and TCGA-LUNG show AMD-MIL's superiority over state-of-the-art methods.