HASD: Hierarchical Adaption for pathology Slide-level Domain-shift

2025-06-30

AI Summary

In pathological AI, whole-slide images (WSIs) exhibit significant domain shift across institutions, yet existing methods predominantly rely on patch-level modeling, failing to capture clinically essential global structural patterns and diagnostic semantics. To address this, we propose HASD—a Hierarchical Adaptive Slice-level framework—enabling the first efficient slice-level domain adaptation for WSIs. HASD jointly aligns features across multiple scales via: (1) domain-level feature alignment, (2) slide-level geometric invariance regularization, (3) patch-level attention consistency constraints, and (4) a lightweight prototype selection mechanism. This design preserves morphological structures and local diagnostic cues while substantially reducing computational overhead. Evaluated on five public datasets across two clinical tasks—HER2 grading and UCEC survival prediction—HASD achieves +4.1% AUROC gain and +3.9% improvement in C-index, respectively, with markedly reduced annotation and computational costs.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Medical and Biological Imaging
• Humans and AI: Other Foundations of Human Computation & AI

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• 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

Abstract

Domain shift is a critical problem for pathology AI as pathology data is heavily influenced by center-specific conditions. Current pathology domain adaptation methods focus on image patches rather than WSI, thus failing to capture global WSI features required in typical clinical scenarios. In this work, we address the challenges of slide-level domain shift by proposing a Hierarchical Adaptation framework for Slide-level Domain-shift (HASD). HASD achieves multi-scale feature consistency and computationally efficient slide-level domain adaptation through two key components: (1) a hierarchical adaptation framework that integrates a Domain-level Alignment Solver for feature alignment, a Slide-level Geometric Invariance Regularization to preserve the morphological structure, and a Patch-level Attention Consistency Regularization to maintain local critical diagnostic cues; and (2) a prototype selection mechanism that reduces computational overhead. We validate our method on two slide-level tasks across five datasets, achieving a 4.1% AUROC improvement in a Breast Cancer HER2 Grading cohort and a 3.9% C-index gain in a UCEC survival prediction cohort. Our method provides a practical and reliable slide-level domain adaption solution for pathology institutions, minimizing both computational and annotation costs.

Problem

• Addresses slide-level domain shift in pathology AI
• Improves multi-scale feature consistency for WSI
• Reduces computational overhead in domain adaptation

Innovation

• Hierarchical adaptation for multi-scale feature consistency
• Domain-level alignment and slide-level geometric invariance
• Prototype selection reduces computational overhead