arXiv cs.AI· Yiheng Xiong, Luisa Gall\'ee, Daniel Santak Wolf, Heiko Hillenhagen, Michael G\"otz·· 15 小时前AI 评分29
医学影像无监督域适应距临床部署还有多远？完整 UDA 流水线评估
How Far from Clinical Deployment? Evaluating the Complete Unsupervised Domain Adaptation Pipeline in Medical Imaging
AI 导读
一项覆盖 9 个医学影像数据集、11 个跨域场景、10 种 UDA 算法与 13 种无标签选择方法的研究，共评估超 80,000 个训练模型，发现通常存在可用的适配模型，但在无目标标签时难以识别。验证器选出的模型与最优模型之间存在较大且结构性的目标域性能差距，且没有一种验证器始终可靠。集成与小规模目标标注预算两种策略能缩小但无法完全消除该差距，研究认为选择环节是推动 UDA 走向临床的关键。
正文 · 原文
Abstract:Deploying unsupervised domain adaptation (UDA) in clinical practice requires choosing which algorithm to use and which of its trained models to ship. However, the deployment (target) domain is unlabeled, so models cannot be evaluated directly on it, leaving it unclear which to select. We address this by evaluating the complete UDA pipeline, considering both adaptation and label-free selection together. Our study covers eleven clinically relevant cross-domain scenarios from nine medical imaging datasets, with ten UDA algorithms and 13 label-free selection methods (validators), evaluating over 80,000 trained models in total. By this, we find that a capable adapted model usually exists, but identifying it without target labels is difficult: the validator-selected models leave a large and structural target performance gap to the best available one, with no evaluated validator consistently reliable. Towards closing it, we explore two strategies, ensembling and a small target-labeling budget; both narrow this gap but do not close it entirely. Overall, deployable UDA depends on the complete pipeline; addressing the less explored selection step could bring much of current UDA closer to clinical use.
| Comments: | Accepted to WACV 2027 |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.12035 [cs.CV] |
| (or arXiv:2608.12035v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2608.12035 arXiv-issued DOI via DataCite |
Submission history
From: Yiheng Xiong [view email]
[v1]
Wed, 12 Aug 2026 13:19:14 UTC (366 KB)
[v2]
Sun, 25 Oct 2026 14:50:25 UTC (374 KB)
来源：arXiv cs.AI · arxiv.org