arXiv：cs.LG· Marco Riedenauer, Daniel Kienzle, Pratik Mayekar, Rainer Lienhart·· 4 小时前AI 评分39
SADUSI 多源超声基准揭示当代自监督异常检测方法的局限
A Multi-Source Ultrasound Benchmark Revealing the Limits of Contemporary Self-Supervised Anomaly Detection Methods
AI 导读
研究者提出 SADUSI 多源超声基准，覆盖多解剖区域、视图与采集协议，用于训练和评估自监督异常检测方法。测试显示，AnoDDPM、DeCo-Diff 等基于重建的扩散方法像素级 AUROC 仅 0.56-0.72、最大 F1 为 0.10-0.26；基于特征的 PatchCore 变体表现更好，AUROC 达 0.76-0.83，但最大 F1 仍只有 0.14-0.40。
正文
Abstract:Self-supervised anomaly detection is a promising paradigm for medical ultrasound, as normal images are often easier to obtain than exhaustive annotations of all possible pathologies. However, most existing evaluations are limited to a single anatomy or task, making it unclear whether models learn a robust notion of normal ultrasound appearance or only a source-specific representation. We introduce the SADUSI benchmark, a multi-source ultrasound dataset designed to train and evaluate anomaly detection methods across a broad range of anatomical regions, views, and acquisition protocols. The goal of SADUSI is to provide a diverse normal ultrasound distribution and a benchmark for visible structural anomalies that can be assessed from single images. We evaluate representative self-supervised anomaly detection methods and find that current approaches struggle in this setting. In particular, reconstruction-based diffusion methods such as AnoDDPM and DeCo-Diff achieve pixel-level AUROC values of 0.56-0.72 and maximum F1 scores of 0.10-0.26, indicating limited separation of pathology from normal image regions. Feature-based PatchCore variants perform better, reaching pixel-level AUROC values of 0.76-0.83, but remain limited with maximum F1 scores of 0.14-0.40. These findings suggest that broad multi-source ultrasound anomaly detection remains an open challenge and that SADUSI can serve as a resource for developing methods that generalize beyond anatomy-specific settings.
| Comments: | 7 pages, 3 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.09677 [cs.CV] |
| (or arXiv:2610.09677v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.09677 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | 2026 IEEE 9th International Conference on Multimedia Information Processing and Retrieval (MIPR), Bangkok, Thailand, 2026, pp. 340-346 |
| Related DOI: | https://doi.org/10.1109/MIPR70517.2026.00061
DOI(s) linking to related resources |
Submission history
From: Marco Riedenauer [view email]
[v1]
Wed, 7 Oct 2026 08:39:51 UTC (1,004 KB)
来源：arXiv：cs.LG · arxiv.org