Do MIL Models Transfer?

2025-06-10

AI Summary

Multi-instance learning (MIL) models in computational pathology face generalization bottlenecks under small-scale, weakly supervised clinical data, and their transferability has lacked systematic evaluation. Method: We conduct the first large-scale, systematic study of cross-organ and cross-task transferability of MIL models—evaluating 11 state-of-the-art MIL architectures across 21 histomorphological and molecular subtype prediction tasks—and establish a standardized MIL implementation framework with an open-source library of pretrained weights. Results: Pretrained MIL models consistently outperform from-scratch baselines; pan-cancer pretraining proves especially effective—achieving superior performance to slide-level foundation models with minimal fine-tuning data and substantially enhancing cross-organ generalization. This work establishes a reproducible, scalable transfer learning paradigm for computational pathology, advancing the clinical deployment of MIL models.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Large Vision Models
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Multiple Instance Learning (MIL) is a cornerstone approach in computational pathology (CPath) for generating clinically meaningful slide-level embeddings from gigapixel tissue images. However, MIL often struggles with small, weakly supervised clinical datasets. In contrast to fields such as NLP and conventional computer vision, where transfer learning is widely used to address data scarcity, the transferability of MIL models remains poorly understood. In this study, we systematically evaluate the transfer learning capabilities of pretrained MIL models by assessing 11 models across 21 pretraining tasks for morphological and molecular subtype prediction. Our results show that pretrained MIL models, even when trained on different organs than the target task, consistently outperform models trained from scratch. Moreover, pretraining on pancancer datasets enables strong generalization across organs and tasks, outperforming slide foundation models while using substantially less pretraining data. These findings highlight the robust adaptability of MIL models and demonstrate the benefits of leveraging transfer learning to boost performance in CPath. Lastly, we provide a resource which standardizes the implementation of MIL models and collection of pretrained model weights on popular CPath tasks, available at https://github.com/mahmoodlab/MIL-Lab

Problem

• Evaluating transfer learning capabilities of MIL models in pathology
• Assessing MIL model performance across diverse organs and tasks
• Standardizing MIL implementation and pretrained weights for computational pathology

Innovation

• Evaluates MIL model transfer learning capabilities
• Pretrained MIL models outperform scratch training
• Pancancer pretraining enhances generalization performance