InfoOOD:信息瓶的优化,以后 hoc 医疗图像外分发检测检测
Brayden Schott1, Žan Klaneček2, Victor Santoro-Fernandes1
1Department of Medical Physics, School of Medicine and Public Health, University of Wisconsin, Madison, WI, United States of America.
Physics in medicine and biology
|October 8, 2025
概括
我们介绍InfoOOD,这是一个信息理论方法,用于检测医疗成像中的分布外 (OOD) 数据. InfoOOD显著改善了对人工制造物诱导的变异的检测,提高了临床环境中深度学习模型的安全性.
科学领域:
- 医疗成像医学成像
- 人工智能的人工智能
- 信息理论 信息理论
背景情况:
- 深度学习模型与分布外 (OOD) 数据扎,导致临床应用中的潜在失败.
- 现有的OOD检测方法对医学成像中常见的微妙变异缺乏灵敏度.
研究的目的:
- 引入和验证InfoOOD,一种新的后期,基于信息的方法,用于检测医疗图像中的OOD数据.
- 与现有的OOD检测方法相比,评估InfoOOD的敏感性和临床相关性.
主要方法:
- 在腹部CT图像 (N=157) 上训练了3D U-Net用于肝脏和病变细分.
- 在测试图像 (N=40) 上模拟基于物理的工件 (低剂量,稀疏视图,环形工件).
- 评估了InfoOOD的检测性能与嵌入式基于特征和基于重建的方法相比.
主要成果:
- 工件模拟显著降低了细分性能,随着工件的大小而恶化.
- 在检测工件诱导的OOD数据方面,InfoOOD的表现始终优于现有的方法 (例如,强环工件的AUC=0.93与0.57).
- InfoOOD与细分绩效指标的负相关性更强,表明可靠性更好.
结论:
- InfoOOD是一种新的,高度敏感和临床相关的医学图像OOD检测方法.
- 这种方法通过可靠地识别数据转移,支持在临床环境中安全部署深度学习模型.
相关概念视频
01:18Steps in Outbreak Investigation
In the ever-evolving field of public health, statistical analysis serves as a cornerstone for understanding and managing disease outbreaks. By leveraging various statistical tools, health professionals can predict potential outbreaks, analyze ongoing situations, and devise effective responses to mitigate impact. For that to happen, there are a few possible stages of the analysis:
01:23Investigation of Disease Outbreaks
Multistate foodborne outbreaks pose significant public health risks and require meticulous investigation to identify sources and implement control measures. The Centers for Disease Control and Prevention (CDC) utilizes a dynamic seven-step process for these investigations, integrating data from laboratories, interviews, and environmental assessments to protect public health.Outbreak Detection: The detection of multistate outbreaks typically begins with PulseNet, the CDC's national laboratory...