The Next Layer: Augmenting Foundation Models with Structure-Preserving and Attention-Guided Learning for Local Patches to Global Context Awareness in Computational Pathology

2025-08-27

AI Summary

In computational pathology, existing foundation models struggle to jointly model global tissue spatial architecture and local diagnostic region context, hindering comprehensive tumor microenvironment characterization. To address this, we propose EAGLE-Net—the first framework that synergistically integrates structure-preserving design with attention guidance. It introduces multi-scale absolute positional encoding, a Top-K neighborhood-aware loss, and a background-suppression loss to generate biologically consistent attention maps. Built upon a multiple-instance learning framework, EAGLE-Net fuses features from heterogeneous foundation models (REMEDIES, Uni-V1, and Uni2-h), jointly optimizing spatial encoding, attention mechanisms, and contrastive loss. On a pan-cancer dataset, EAGLE-Net achieves up to a 3% improvement in classification accuracy, attains the highest C-index for survival prediction in 6 out of 7 cancer types, significantly reduces false positives, and precisely localizes clinically critical regions—including tumor invasion fronts, necrotic zones, and immune-infiltrated areas.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Multiagent Systems: Other Foundations of Multi Agent Systems

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Foundation models have recently emerged as powerful feature extractors in computational pathology, yet they typically omit mechanisms for leveraging the global spatial structure of tissues and the local contextual relationships among diagnostically relevant regions - key elements for understanding the tumor microenvironment. Multiple instance learning (MIL) remains an essential next step following foundation model, designing a framework to aggregate patch-level features into slide-level predictions. We present EAGLE-Net, a structure-preserving, attention-guided MIL architecture designed to augment prediction and interpretability. EAGLE-Net integrates multi-scale absolute spatial encoding to capture global tissue architecture, a top-K neighborhood-aware loss to focus attention on local microenvironments, and background suppression loss to minimize false positives. We benchmarked EAGLE-Net on large pan-cancer datasets, including three cancer types for classification (10,260 slides) and seven cancer types for survival prediction (4,172 slides), using three distinct histology foundation backbones (REMEDIES, Uni-V1, Uni2-h). Across tasks, EAGLE-Net achieved up to 3% higher classification accuracy and the top concordance indices in 6 of 7 cancer types, producing smooth, biologically coherent attention maps that aligned with expert annotations and highlighted invasive fronts, necrosis, and immune infiltration. These results position EAGLE-Net as a generalizable, interpretable framework that complements foundation models, enabling improved biomarker discovery, prognostic modeling, and clinical decision support

Problem

• Augmenting foundation models to capture global tissue spatial structure
• Designing attention-guided MIL framework for slide-level predictions
• Improving interpretability and accuracy in computational pathology tasks

Innovation

• Structure-preserving attention-guided MIL architecture
• Multi-scale absolute spatial encoding for global tissue
• Top-K neighborhood-aware loss for local microenvironments