Unsupervised Domain Adaptation via Content Alignment for Hippocampus Segmentation

2025-10-14

AI Summary

Medical image segmentation models suffer from degraded generalizability when deployed across datasets due to dual domain shifts in MRI—both in style (appearance) and content (anatomical structure)—particularly for fine-grained structures such as the hippocampus. This paper proposes an unsupervised domain adaptation framework that jointly optimizes a bidirectional deformable image registration (DIR) module and a segmentation-discriminator network to achieve anatomically plausible content alignment. Style harmonization is achieved via z-score normalization, and the segmentation, discriminator, and DIR modules are trained end-to-end in a unified manner. Evaluated on multi-center hippocampal segmentation from MRI, the method substantially mitigates content shift: in cross-domain transfer from healthy young subjects to dementia patients, it achieves up to a 15% improvement in Dice score over source-only baselines, outperforming state-of-the-art data augmentation and domain adaptation approaches.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Segmentation
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Deep learning models for medical image segmentation often struggle when deployed across different datasets due to domain shifts - variations in both image appearance, known as style, and population-dependent anatomical characteristics, referred to as content. This paper presents a novel unsupervised domain adaptation framework that directly addresses domain shifts encountered in cross-domain hippocampus segmentation from MRI, with specific emphasis on content variations. Our approach combines efficient style harmonisation through z-normalisation with a bidirectional deformable image registration (DIR) strategy. The DIR network is jointly trained with segmentation and discriminator networks to guide the registration with respect to a region of interest and generate anatomically plausible transformations that align source images to the target domain. We validate our approach through comprehensive evaluations on both a synthetic dataset using Morpho-MNIST (for controlled validation of core principles) and three MRI hippocampus datasets representing populations with varying degrees of atrophy. Across all experiments, our method outperforms existing baselines. For hippocampus segmentation, when transferring from young, healthy populations to clinical dementia patients, our framework achieves up to 15% relative improvement in Dice score compared to standard augmentation methods, with the largest gains observed in scenarios with substantial content shift. These results highlight the efficacy of our approach for accurate hippocampus segmentation across diverse populations.

Problem

• Addressing domain shifts in cross-dataset medical image segmentation
• Aligning anatomical content variations for hippocampus MRI segmentation
• Improving unsupervised adaptation from healthy to dementia patient scans

Innovation

• Bidirectional deformable image registration aligns anatomical content
• Joint training integrates segmentation with adversarial domain adaptation
• Style harmonization via z-normalization addresses appearance variations