HypCBC: Domain-Invariant Hyperbolic Cross-Branch Consistency for Generalizable Medical Image Analysis

2026-02-03

AI Summary

This work addresses the challenge of distribution shifts in medical image analysis caused by variations in imaging devices, protocols, and patient populations. To enhance model generalizability, the authors propose a novel domain generalization method grounded in hyperbolic geometry, embedding a Vision Transformer within a hyperbolic space to better capture the hierarchical structure inherent in medical images. The approach introduces the first unsupervised domain-invariant cross-branch consistency constraint, enabling the learning of robust and transferable feature representations. This study presents the first systematic validation of hyperbolic representation learning for medical image analysis. Extensive experiments demonstrate that the proposed method significantly outperforms existing Euclidean-based approaches across 11 in-distribution datasets and three domain generalization benchmarks, achieving an average AUC improvement of 2.1% across multimodal and multiscale tasks.

Technology Category

• Computer Vision: Medical and Biological Imaging
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Robust generalization beyond training distributions remains a critical challenge for deep neural networks. This is especially pronounced in medical image analysis, where data is often scarce and covariate shifts arise from different hardware devices, imaging protocols, and heterogeneous patient populations. These factors collectively hinder reliable performance and slow down clinical adoption. Despite recent progress, existing learning paradigms primarily rely on the Euclidean manifold, whose flat geometry fails to capture the complex, hierarchical structures present in clinical data. In this work, we exploit the advantages of hyperbolic manifolds to model complex data characteristics. We present the first comprehensive validation of hyperbolic representation learning for medical image analysis and demonstrate statistically significant gains across eleven in-distribution datasets and three ViT models. We further propose an unsupervised, domain-invariant hyperbolic cross-branch consistency constraint. Extensive experiments confirm that our proposed method promotes domain-invariant features and outperforms state-of-the-art Euclidean methods by an average of $+2.1\%$ AUC on three domain generalization benchmarks: Fitzpatrick17k, Camelyon17-WILDS, and a cross-dataset setup for retinal imaging. These datasets span different imaging modalities, data sizes, and label granularities, confirming generalization capabilities across substantially different conditions. The code is available at https://github.com/francescodisalvo05/hyperbolic-cross-branch-consistency .

Problem

• domain generalization
• medical image analysis
• covariate shift
• data scarcity
• distribution shift

Innovation

• hyperbolic representation learning
• domain generalization
• cross-branch consistency
• medical image analysis
• domain-invariant features