Self-Supervision Enhances Instance-based Multiple Instance Learning Methods in Digital Pathology: A Benchmark Study

2025-05-02

AI Summary

Instance-level multiple instance learning (MIL) for whole-slide image (WSI) classification in digital pathology faces a trade-off between high performance and clinical interpretability. Method: We systematically investigate the impact of self-supervised learning (SSL) on instance-level MIL, proposing the first unified MIL benchmark tailored to pathology—incorporating six SSL methods, four backbone architectures, four base models, and pathology-specific adaptations—and introducing four novel instance-level MIL approaches. Contribution/Results: Extensive experiments (710 configurations) on BRACS and Camelyon16 demonstrate that high-quality SSL features substantially boost lightweight instance-level MIL, surpassing state-of-the-art embedding-based MIL methods while preserving superior lesion localization interpretability. Our findings establish that, under robust SSL representations, simple and transparent instance-level modeling offers greater clinical deployability than complex, black-box alternatives.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Interpretability, Explainability, and Transparency
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Multiple Instance Learning (MIL) has emerged as the best solution for Whole Slide Image (WSI) classification. It consists of dividing each slide into patches, which are treated as a bag of instances labeled with a global label. MIL includes two main approaches: instance-based and embedding-based. In the former, each patch is classified independently, and then the patch scores are aggregated to predict the bag label. In the latter, bag classification is performed after aggregating patch embeddings. Even if instance-based methods are naturally more interpretable, embedding-based MILs have usually been preferred in the past due to their robustness to poor feature extractors. However, recently, the quality of feature embeddings has drastically increased using self-supervised learning (SSL). Nevertheless, many authors continue to endorse the superiority of embedding-based MIL. To investigate this further, we conduct 710 experiments across 4 datasets, comparing 10 MIL strategies, 6 self-supervised methods with 4 backbones, 4 foundation models, and various pathology-adapted techniques. Furthermore, we introduce 4 instance-based MIL methods never used before in the pathology domain. Through these extensive experiments, we show that with a good SSL feature extractor, simple instance-based MILs, with very few parameters, obtain similar or better performance than complex, state-of-the-art (SOTA) embedding-based MIL methods, setting new SOTA results on the BRACS and Camelyon16 datasets. Since simple instance-based MIL methods are naturally more interpretable and explainable to clinicians, our results suggest that more effort should be put into well-adapted SSL methods for WSI rather than into complex embedding-based MIL methods.

Problem

• Evaluating self-supervised learning for instance-based MIL in pathology
• Comparing performance of instance-based vs embedding-based MIL methods
• Improving interpretability of WSI classification with SSL-enhanced MIL

Innovation

• Uses self-supervised learning for feature extraction
• Introduces new instance-based MIL methods
• Compares multiple SSL and MIL strategies