分布偏移下基于伪标签差异的医学图像分类无基准模型选择
Label-Free Foundational Model Selection for Medical Image Classification under Distribution Shift via Pseudo Label Discrepancy
浏览论文内容
中文总结 AI 辅助
针对分布偏移下医学图像分类的无基准模型选择问题，提出基于SUDO框架的AURCC准则，在胸部X射线分类任务中实现了与真实排名高相关性的模型排名，小源标注数据下表现更优。
中文摘要 AI 辅助
基础模型越来越多地被部署用于医学图像分析，但在部署时典型的机构间分布偏移场景下，它们的性能差异很大，且在目标域标签极少可用的情况下无法得知其性能，这留下了一个未解决的实际问题：给定多个候选基础模型和源域的标注数据，应选择哪一个部署到无标注的目标域？我们提出一种无基准选择准则，该准则基于SUDO框架（一种无需真实标注即可评估临床AI系统的框架）。SUDO通过预测概率划分无标注的目标数据，并针对每个区域测量反映类别污染的伪标签差异；跨区域聚合后得到一个既不需要目标标注也不需要微调的分数（AURCC）。我们在三种跨医院偏移场景下的胸部X射线分类任务中，于零样本和MLP探测 regime下，展示了AURCC可用于对多种视觉语言模型（BioMedCLIP、CXR-CLIP、CheXzero、MedCLIP、MedImageInsight、CLIP）进行排名。AURCC排名与真实排名的Spearman相关系数最高达0.943（p<0.05）。与通过保留源域准确率进行排名的自然基线相比，当标注源域数据量较大时AURCC具有竞争力，而当标注源域数据量较小时，AURCC能产生更准确的排名——这是资源受限场景下的关注 regime。
英文摘要
Foundation models are increasingly deployed for medical image analysis. However, under the inter-institutional distribution shift typical of deployment, their performance varies widely and cannot be known without target-domain labels, which are rarely available. This leaves a practical question unresolved: given several candidate foundational models and labeled-data from a source domain, which one to deploy in an unlabeled target domain? We propose a label-free selection criterion built on SUDO, a framework for evaluating clinical AI systems without ground-truth annotations. SUDO partitions the unlabeled target data by predicted probability and, for each region, measures a pseudo-label discrepancy reflecting class contamination; aggregated across regions, this yields a score (AURCC) requiring neither target annotation nor fine-tuning. We show that AURCC can be used to rank a variety of vision-language models (BioMedCLIP, CXR-CLIP, CheXzero, MedCLIP, MedImageInsight, CLIP) on chest X-ray classification across three inter-hospital shift scenarios, under zero-shot and MLP-probe regimes. The AURCC ranking recovers the ground-truth ranking with Spearman rho up to 0.943 (p<0.05). Against the natural baseline of ranking by held-out source accuracy, AURCC is competitive when the labeled source is large and yields a more accurate ranking once it is small; the regime of interest in resource-constrained settings.
发表机构
- Universidad de Buenos Aires（布宜诺斯艾利斯大学）
- CONICET（阿根廷国家科学技术研究委员会）
- Universidad Nacional del Litoral（国立 littoral 大学）
- Research Institute for Signals, Systems and Computational Intelligence（信号、系统与计算智能研究所）
机构由 AI 辅助整理，请以论文原文为准。
