UWarp: A Whole Slide Image Registration Pipeline to Characterize Scanner-Induced Local Domain Shift

2025-03-26

AI Summary

This work addresses patch-level domain shift induced by heterogeneous digital slide scanners in computational pathology. We propose the first local domain shift quantification framework specifically designed for computational pathology. Methodologically, we introduce a hierarchical registration architecture: global affine correction followed by local non-rigid deformation compensation, with registration accuracy quantitatively evaluated via target registration error (TRE). On two private datasets, median TRE is <4 pixels (<1 μm at 40× magnification), achieving significant efficiency gains. A key finding is a strong spatial correlation between model prediction variability and local tissue density. To our knowledge, this is the first study to enable interpretable, patch-level quantification of domain shift. Our framework establishes a novel paradigm and provides a practical tool for mitigating scanner-induced domain shift in digital pathology.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Interpretability, Explainability, and Transparency
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Large-scale security measurements

Abstract

Histopathology slide digitization introduces scanner-induced domain shift that can significantly impact computational pathology models based on deep learning methods. In the state-of-the-art, this shift is often characterized at a broad scale (slide-level or dataset-level) but not patch-level, which limits our comprehension of the impact of localized tissue characteristics on the accuracy of the deep learning models. To address this challenge, we present a domain shift analysis framework based on UWarp, a novel registration tool designed to accurately align histological slides scanned under varying conditions. UWarp employs a hierarchical registration approach, combining global affine transformations with fine-grained local corrections to achieve robust tissue patch alignment. We evaluate UWarp using two private datasets, CypathLung and BosomShieldBreast, containing whole slide images scanned by multiple devices. Our experiments demonstrate that UWarp outperforms existing open-source registration methods, achieving a median target registration error (TRE) of less than 4 pixels (<1 micrometer at 40x magnification) while significantly reducing computational time. Additionally, we apply UWarp to characterize scanner-induced local domain shift in the predictions of Breast-NEOprAIdict, a deep learning model for breast cancer pathological response prediction. We find that prediction variability is strongly correlated with tissue density on a given patch. Our findings highlight the importance of localized domain shift analysis and suggest that UWarp can serve as a valuable tool for improving model robustness and domain adaptation strategies in computational pathology.

Problem

• Analyze scanner-induced domain shift in histopathology slide digitization
• Address patch-level domain shift impact on deep learning models
• Improve registration accuracy for multi-scanner whole slide images

Innovation

• Hierarchical registration combining global and local corrections
• Achieves sub-micrometer alignment accuracy at 40x magnification
• Characterizes scanner-induced domain shift at patch-level