Robust sensitivity control in digital pathology via tile score distribution matching

2025-02-27

AI Summary

Digital pathology models often suffer from uncontrolled sensitivity degradation across multi-center deployments due to distributional shift, failing to meet stringent clinical requirements on critical performance metrics. To address this, we propose a sensitivity-robust calibration method that synergistically integrates optimal transport (OT) and multiple-instance learning (MIL). By explicitly modeling and aligning slice-level score distributions, our approach enables precise sensitivity constraint enforcement for whole-slide image (WSI) classification models using only 5–10 calibrated WSIs. Unlike conventional domain generalization paradigms centered on AUC optimization, ours is the first to incorporate OT into the MIL framework to achieve sensitivity-controllable generalization. Extensive validation across multiple real-world cohorts and diagnostic tasks demonstrates stable sensitivity error below 2%, significantly enhancing clinical deployment reliability and regulatory compliance.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Distributed Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Deploying digital pathology models across medical centers is challenging due to distribution shifts. Recent advances in domain generalization improve model transferability in terms of aggregated performance measured by the Area Under Curve (AUC). However, clinical regulations often require to control the transferability of other metrics, such as prescribed sensitivity levels. We introduce a novel approach to control the sensitivity of whole slide image (WSI) classification models, based on optimal transport and Multiple Instance Learning (MIL). Validated across multiple cohorts and tasks, our method enables robust sensitivity control with only a handful of calibration samples, providing a practical solution for reliable deployment of computational pathology systems.

Problem

• Control sensitivity in digital pathology models
• Address distribution shifts in medical centers
• Ensure robust model transferability with minimal calibration

Innovation

• Optimal transport for sensitivity control
• Multiple Instance Learning in pathology
• Calibration samples for reliable deployment