One Size does not Fit All: Heterogeneous Latent Space Alignment for Unsupervised Domain Adaptation

2026-06-19

AI Summary

This work addresses the performance degradation in medical image segmentation caused by domain shift by proposing the ADualVUOT framework, which overcomes the limitations of existing unsupervised domain adaptation methods—namely, weak latent representation capacity and reliance on static augmentation. The proposed approach integrates a dual-encoder variational autoencoder with continuous normalizing flows to enhance posterior expressiveness and, for the first time, combines unbalanced optimal transport with the Gaussian Gromov–Wasserstein distance to achieve structure-aware alignment of heterogeneous latent spaces. Additionally, it introduces a dynamic adversarial augmentation strategy that synthesizes worst-case samples to improve model robustness. Experimental results demonstrate that ADualVUOT significantly outperforms current optimal transport–based methods across multiple medical image segmentation benchmarks, effectively enhancing cross-domain generalization.

Abstract

Domain shift remains a major obstacle to the reliable deployment of machine learning models in high-stakes environments such as healthcare. While Domain adaptation aims to mitigate these effects, existing approaches suffer from limited expressiveness of latent representations and a reliance on handcrafted, static augmentations. In this work, we address these limitations by proposing a novel deep learning architecture for Unsupervised Domain Adaptation (UDA), specifically optimized for medical image segmentation. Our framework, ADualVUOT, integrates a dual-encoder Variational Autoencoder (VAE) with Continuous Normalizing Flows (CNFs) to increase modeling flexibility and posterior expressiveness. To achieve domain alignment, we leverage Unbalanced Optimal Transport (UOT) through the Gaussian-Gromov-Wasserstein (GGW) distance, which handles structural and topological discrepancies between domains. Furthermore, we incorporate an adversarial augmentation scheme to synthesize worst-case compositions, thus enhancing model robustness. Extensive experiments on medical imaging benchmarks show significant gains over prior OT-based approaches.

Problem

• Domain shift
• Unsupervised Domain Adaptation
• Latent space alignment
• Medical image segmentation
• Representation expressiveness

Innovation

• Unsupervised Domain Adaptation
• Variational Autoencoder
• Continuous Normalizing Flows
• Unbalanced Optimal Transport
• Adversarial Augmentation