Foundation Models -- A Panacea for Artificial Intelligence in Pathology?

2025-02-28

AI Summary

Whether foundation models (FMs) serve as a “universal solution” for AI in digital pathology—specifically for prostate cancer diagnosis and Gleason grading—remains unclear. Method: We constructed a standardized, multi-center evaluation framework using >100,000 needle biopsy images from 15 institutions across 11 countries, comparing self-supervised FMs against task-specific convolutional neural networks (TS-CNNs) trained end-to-end. Contribution/Results: Contrary to prevailing assumptions, FMs showed no statistically significant advantage over TS-CNNs under data-sufficient conditions. TS-CNNs matched or surpassed FMs in diagnostic accuracy, Gleason grade concordance, and critical misclassification rates, while demonstrating superior cross-device generalizability and consuming only ~2.9% of the FM’s inference energy. These findings empirically validate that lightweight, task-optimized, robust architectures—not generic FMs—are more clinically deployable for routine pathology AI applications.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Other Foundations of Computer Vision
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

The role of artificial intelligence (AI) in pathology has evolved from aiding diagnostics to uncovering predictive morphological patterns in whole slide images (WSIs). Recently, foundation models (FMs) leveraging self-supervised pre-training have been widely advocated as a universal solution for diverse downstream tasks. However, open questions remain about their clinical applicability and generalization advantages over end-to-end learning using task-specific (TS) models. Here, we focused on AI with clinical-grade performance for prostate cancer diagnosis and Gleason grading. We present the largest validation of AI for this task, using over 100,000 core needle biopsies from 7,342 patients across 15 sites in 11 countries. We compared two FMs with a fully end-to-end TS model in a multiple instance learning framework. Our findings challenge assumptions that FMs universally outperform TS models. While FMs demonstrated utility in data-scarce scenarios, their performance converged with - and was in some cases surpassed by - TS models when sufficient labeled training data were available. Notably, extensive task-specific training markedly reduced clinically significant misgrading, misdiagnosis of challenging morphologies, and variability across different WSI scanners. Additionally, FMs used up to 35 times more energy than the TS model, raising concerns about their sustainability. Our results underscore that while FMs offer clear advantages for rapid prototyping and research, their role as a universal solution for clinically applicable medical AI remains uncertain. For high-stakes clinical applications, rigorous validation and consideration of task-specific training remain critically important. We advocate for integrating the strengths of FMs and end-to-end learning to achieve robust and resource-efficient AI pathology solutions fit for clinical use.

Problem

• Evaluating clinical applicability of foundation models in pathology
• Comparing performance of foundation models vs. task-specific models
• Assessing energy efficiency and sustainability of foundation models

Innovation

• Foundation models for diverse pathology tasks
• Comparison with task-specific end-to-end models
• Energy efficiency and clinical validation importance