arXiv cs.CL· Soroosh Tayebi Arasteh, Mahshad Lotfinia, Sven Nebelung, Daniel Truhn·· 12 小时前AI 评分36
选择性关联精度：评估已发布医疗视觉语言模型的隐私风险
Selective Linkage Precision for Released Medical Vision-Language Models
AI 导读
研究提出“选择性关联精度”指标，衡量按性别、年龄段、投照体位等属性过滤候选报告后，视觉语言模型在最高置信度查询中正确匹配影像与报告的比例。在 ReXGradient-160K 上，MedSigLIP-448 在预算 100 时精度达 67.68%，而仅用属性策略为 24.99%、仅用模型为 1.98%；加入采集年份后精度升至 99.05%。该指标无需重训练即可推理，可帮助医院评估应公开哪些属性。
正文 · 原文
Abstract:Hospitals may release de-identified chest radiographs while restricting access to their reports. An attacker who already has reports from the same population and their metadata can use a vision-language model (VLM) to match images to reports. We introduce selective linkage precision, which measures correct matches among the highest-confidence queries after filtering candidates by the available attributes. Across 8 public VLMs and 3 chest radiograph datasets, the maximum whole-pool Recall@1 is 0.40%. However, this metric averages over every query without attribute filtering. On ReXGradient-160K, MedSigLIP-448 reaches 67.68% precision at a budget of 100 after filtering by sex, age band, and projection, with a median of 327 candidate reports. On the same eligible query set, an attribute-only policy reaches 24.99% and the model alone reaches 1.98%. Adding acquisition year, available in only 1 of the 3 datasets, yields 99.05% precision and a median of 24 candidates. Uniform guessing on the observed selected queries has an expected precision of 19.91%. Expanding the primary-setting collection to all 123,100 ReXGradient studies lowers precision to 37.23%. On BioViL-T, the tested head finetuning procedures do not lower linkage below the public checkpoint. Selective linkage precision requires inference without retraining and could help hospitals evaluate which attributes to publish.
| Comments: | Accepted at ML4H 2026 Proceedings, Sydney, Australia |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2606.02276 [cs.CV] |
| (or arXiv:2606.02276v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.02276 arXiv-issued DOI via DataCite |
Submission history
From: Soroosh Tayebi Arasteh [view email]
[v1]
Mon, 1 Jun 2026 14:01:46 UTC (366 KB)
[v2]
Sat, 31 Oct 2026 12:38:24 UTC (191 KB)
来源：arXiv cs.CL · arxiv.org