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

TL;DR

This paper introduces HypCBC, a hyperbolic manifold-based method for medical image analysis that enhances domain generalization and outperforms Euclidean approaches across multiple datasets and models.

Contribution

It presents the first validation of hyperbolic representation learning in medical imaging and proposes an unsupervised, domain-invariant hyperbolic cross-branch consistency constraint.

Findings

Significant improvements in AUC across 11 datasets and 3 ViT models.

Outperforms state-of-the-art Euclidean methods by +2.1% AUC.

Effective across diverse imaging modalities and data conditions.

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 .