Unsupervised Domain Adaptation via Similarity-based Prototypes for Cross-Modality Segmentation

2025-10-23

AI Summary

To address semantic shift and category missing in unsupervised domain adaptation (UDA) for cross-modality medical image segmentation, this paper proposes a similarity-constrained class-prototype learning framework. Methodologically, it constructs an updateable prototype dictionary to explicitly model source-domain class prototypes in deep embedding space, and employs similarity-driven prototype alignment alongside prototype-level contrastive learning to narrow the semantic gap between source and target domains while mitigating category collapse induced by target-domain label absence. The key innovation lies in the organic integration of prototype representation, dynamic dictionary updating, and cross-domain contrastive learning—enabling robust, annotation-free segmentation on the target domain. Extensive experiments on multiple cross-modality benchmarks (e.g., MRI→CT, T1→T2) demonstrate consistent and significant improvements over state-of-the-art methods, achieving average Dice score gains of 3.2–5.8 percentage points.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Deep learning models have achieved great success on various vision challenges, but a well-trained model would face drastic performance degradation when applied to unseen data. Since the model is sensitive to domain shift, unsupervised domain adaptation attempts to reduce the domain gap and avoid costly annotation of unseen domains. This paper proposes a novel framework for cross-modality segmentation via similarity-based prototypes. In specific, we learn class-wise prototypes within an embedding space, then introduce a similarity constraint to make these prototypes representative for each semantic class while separable from different classes. Moreover, we use dictionaries to store prototypes extracted from different images, which prevents the class-missing problem and enables the contrastive learning of prototypes, and further improves performance. Extensive experiments show that our method achieves better results than other state-of-the-art methods.

Problem

• Addresses domain shift in cross-modality medical image segmentation
• Learns separable class prototypes via similarity constraints
• Uses dictionary-based contrastive learning to prevent class-missing issues

Innovation

• Uses similarity-based prototypes for domain adaptation
• Learns class-wise prototypes in embedding space
• Employs dictionaries for contrastive prototype learning