GAS-MIL: Group-Aggregative Selection Multi-Instance Learning for Ensemble of Foundation Models in Digital Pathology Image Analysis

2025-10-03

AI Summary

In digital pathology image analysis, feature fusion from multiple foundation models (FMs) often relies on manual selection or extensive task-specific fine-tuning, compromising both complementarity and generalizability. To address this, we propose GAS-MIL, a novel framework introducing Group-wise Aggregation and Selection (GAS)—the first mechanism enabling automatic identification and integration of complementary discriminative features from heterogeneous FMs under the multiple-instance learning (MIL) paradigm, without fine-tuning any FM or introducing additional learnable parameters. GAS-MIL performs end-to-end integration of multi-source FM outputs, significantly reducing ensemble complexity. Evaluated on three real-world histopathological datasets—prostate, ovarian, and breast cancer—GAS-MIL matches or surpasses the performance of individual FMs and conventional MIL methods, demonstrating superior robustness and cross-cancer generalization capability.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

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: Federated recommendation systems and personalization

Abstract

Foundation models (FMs) have transformed computational pathology by providing powerful, general-purpose feature extractors. However, adapting and benchmarking individual FMs for specific diagnostic tasks is often time-consuming and resource-intensive, especially given their scale and diversity. To address this challenge, we introduce Group-Aggregative Selection Multi-Instance Learning (GAS-MIL), a flexible ensemble framework that seamlessly integrates features from multiple FMs, preserving their complementary strengths without requiring manual feature selection or extensive task-specific fine-tuning. Across classification tasks in three cancer datasets-prostate (PANDA), ovarian (UBC-OCEAN), and breast (TCGA-BrCa)-GAS-MIL consistently achieves superior or on-par performance relative to individual FMs and established MIL methods, demonstrating its robustness and generalizability. By enabling efficient integration of heterogeneous FMs, GAS-MIL streamlines model deployment for pathology and provides a scalable foundation for future multimodal and precision oncology applications.

Problem

• Integrating multiple foundation models for digital pathology analysis
• Reducing manual feature selection and fine-tuning requirements
• Achieving robust performance across diverse cancer classification tasks

Innovation

• GAS-MIL integrates multiple foundation models automatically
• It eliminates manual feature selection and fine-tuning
• The framework preserves complementary strengths of diverse models