Robust sensitivity control in digital pathology via tile score distribution matching

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pignet, Arthur, Klein, John, Robin, Genevieve, Olivier, Antoine
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912498769199104
author Pignet, Arthur
Klein, John
Robin, Genevieve
Olivier, Antoine
author_facet Pignet, Arthur
Klein, John
Robin, Genevieve
Olivier, Antoine
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2502_20144
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Robust sensitivity control in digital pathology via tile score distribution matching
Pignet, Arthur
Klein, John
Robin, Genevieve
Olivier, Antoine
Computer Vision and Pattern Recognition
Machine Learning
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.
title Robust sensitivity control in digital pathology via tile score distribution matching
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2502.20144