Attention-based Generative Latent Replay: A Continual Learning Approach for WSI Analysis

2025-05-13

AI Summary

Addressing the dual challenges of catastrophic forgetting and domain shift—across organs, diseases, and institutions—in whole-slide image (WSI) domain-incremental classification, this work proposes a privacy-preserving buffer-free continual learning framework. Our core innovation is an attention-driven generative latent replay mechanism: leveraging a Gaussian Mixture Model (GMM) in the latent space, it synthesizes discriminative WSI instance representations and patch-count distributions without storing raw images or maintaining a replay buffer. Integrated within a multiple-instance learning (MIL) framework, the method enables efficient and secure knowledge updating. Evaluated on multiple multi-center clinical datasets, our approach achieves performance on par with buffered baselines for biomarker detection and molecular status prediction, significantly outperforming other buffer-free methods. It demonstrates strong long-term knowledge retention while ensuring clinical deployability and data privacy compliance.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Medical and Biological Imaging
• Multiagent Systems: Multiagent Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Whole slide image (WSI) classification has emerged as a powerful tool in computational pathology, but remains constrained by domain shifts, e.g., due to different organs, diseases, or institution-specific variations. To address this challenge, we propose an Attention-based Generative Latent Replay Continual Learning framework (AGLR-CL), in a multiple instance learning (MIL) setup for domain incremental WSI classification. Our method employs Gaussian Mixture Models (GMMs) to synthesize WSI representations and patch count distributions, preserving knowledge of past domains without explicitly storing original data. A novel attention-based filtering step focuses on the most salient patch embeddings, ensuring high-quality synthetic samples. This privacy-aware strategy obviates the need for replay buffers and outperforms other buffer-free counterparts while matching the performance of buffer-based solutions. We validate AGLR-CL on clinically relevant biomarker detection and molecular status prediction across multiple public datasets with diverse centers, organs, and patient cohorts. Experimental results confirm its ability to retain prior knowledge and adapt to new domains, offering an effective, privacy-preserving avenue for domain incremental continual learning in WSI classification.

Problem

• Addresses domain shifts in WSI classification across different organs and institutions
• Proposes continual learning without storing original data to preserve privacy
• Ensures high-quality synthetic samples via attention-based filtering of patch embeddings

Innovation

• Attention-based filtering for salient patch embeddings
• Gaussian Mixture Models synthesize WSI representations
• Privacy-aware continual learning without replay buffers