The best performance in the CARE 2025 -- Liver Task (LiSeg-Contrast): Contrast-Aware Semi-Supervised Segmentation with Domain Generalization and Test-Time Adaptation

2025-10-05

AI Summary

To address annotation scarcity, heterogeneous scanning protocols, and cross-center domain shifts in liver segmentation from contrast-enhanced MRI, this paper proposes CoSSeg-TTA—a test-time adaptive framework built upon nnU-Netv2. It integrates a semi-supervised mean-teacher mechanism with a domain adaptation module, and introduces three key innovations: (i) randomized histogram-based style transfer to mitigate scanner-induced appearance variation; (ii) a learnable contrast-aware network to enhance lesion and boundary discrimination; and (iii) continual test-time adaptation for dynamic model updating during inference. Evaluated under extremely low-label settings, CoSSeg-TTA achieves an average Dice score improvement of 3.2% and a 18.7% reduction in Hausdorff distance across multi-center datasets, significantly outperforming the nnU-Netv2 baseline. These results demonstrate superior generalizability to unseen domains and strong clinical applicability.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Accurate liver segmentation from contrast-enhanced MRI is essential for diagnosis, treatment planning, and disease monitoring. However, it remains challenging due to limited annotated data, heterogeneous enhancement protocols, and significant domain shifts across scanners and institutions. Traditional image-to-image translation frameworks have made great progress in domain generalization, but their application is not straightforward. For example, Pix2Pix requires image registration, and cycle-GAN cannot be integrated seamlessly into segmentation pipelines. Meanwhile, these methods are originally used to deal with cross-modality scenarios, and often introduce structural distortions and suffer from unstable training, which may pose drawbacks in our single-modality scenario. To address these challenges, we propose CoSSeg-TTA, a compact segmentation framework for the GED4 (Gd-EOB-DTPA enhanced hepatobiliary phase MRI) modality built upon nnU-Netv2 and enhanced with a semi-supervised mean teacher scheme to exploit large amounts of unlabeled volumes. A domain adaptation module, incorporating a randomized histogram-based style appearance transfer function and a trainable contrast-aware network, enriches domain diversity and mitigates cross-center variability. Furthermore, a continual test-time adaptation strategy is employed to improve robustness during inference. Extensive experiments demonstrate that our framework consistently outperforms the nnU-Netv2 baseline, achieving superior Dice score and Hausdorff Distance while exhibiting strong generalization to unseen domains under low-annotation conditions.

Problem

• Accurate liver segmentation from contrast-enhanced MRI faces limited annotated data challenges
• Heterogeneous enhancement protocols cause domain shifts across scanners and institutions
• Existing methods introduce structural distortions and unstable training in single-modality scenarios

Innovation

• Semi-supervised mean teacher scheme exploits unlabeled volumes
• Domain adaptation module enriches diversity and reduces variability
• Continual test-time adaptation strategy enhances inference robustness