AINet: Anchor Instances Learning for Regional Heterogeneity in Whole Slide Image

TL;DR

AINet introduces anchor instances to effectively address regional heterogeneity in whole slide image analysis, improving representation quality and model performance with fewer computational resources.

Contribution

The paper proposes a novel anchor instance concept and dual modules DAM and ARC, enhancing multi-instance learning for WSI by capturing regional diversity and discriminative features.

Findings

AINet outperforms state-of-the-art methods in WSI analysis.

The framework reduces computational complexity significantly.

DAM and ARC modules improve regional representation quality.

Abstract

Recent advances in multi-instance learning (MIL) have witnessed impressive performance in whole slide image (WSI) analysis. However, the inherent sparsity of tumors and their morphological diversity lead to obvious heterogeneity across regions, posing significant challenges in aggregating high-quality and discriminative representations. To address this, we introduce a novel concept of anchor instance (AI), a compact subset of instances that are representative within their regions (local) and discriminative at the bag (global) level. These AIs act as semantic references to guide interactions across regions, correcting non-discriminative patterns while preserving regional diversity. Specifically, we propose a dual-level anchor mining (DAM) module to \textbf{select} AIs from massive instances, where the most informative AI in each region is extracted by assessing its similarity to both local and global embeddings. Furthermore, to ensure completeness and diversity, we devise an anchor-guided region correction (ARC) module that explores the complementary information from all regions to \textbf{correct} each regional representation. Building upon DAM and ARC, we develop a concise yet effective framework, AINet, which employs a simple predictor and surpasses state-of-the-art methods with substantially fewer FLOPs and parameters. Moreover, both DAM and ARC are modular and can be seamlessly integrated into existing MIL frameworks, consistently improving their performance.