Scholarlink

Image Anomaly Detection (IAD) helps ensure quality by detecting defects in industrial applications, while in healthcare, it aids diagnosis and improves patient outcomes. Recent advances in deep learning and computer vision have significantly improved IAD accuracy and efficiency [30, 40, 57, 58]. Medical Anomaly Detection (AD), used in computer-aided diagnosis (CAD), is more challenging due to sparse data, privacy concerns, and complex imagery [15, 53]. While ML algorithms have been used to detect lung diseases from chest X-rays [5], others [45] note that foreign particle infiltration (a symptom of Tuberculosis) is harder to identify. This review compares state-of-the-art (SOTA) IAD methods, focusing on sample datasets from industrial applications (MVTec AD dataset [8]) and medical applications (NIH Chest X-ray dataset [54]). We identify opportunities to adapt industrial IAD advancements to improve medical IAD, particularly for detecting lung infiltration, potentially enhancing the diagnostic utility of common chest X-rays.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Applying Best Practices from Image Anomaly Detection to Identifying Infiltration in Chest X-Rays: A Review

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Matthew Winchester,
  • Patrick Husson,
  • Xiaodong Qu

摘要

Image Anomaly Detection (IAD) helps ensure quality by detecting defects in industrial applications, while in healthcare, it aids diagnosis and improves patient outcomes. Recent advances in deep learning and computer vision have significantly improved IAD accuracy and efficiency [30, 40, 57, 58]. Medical Anomaly Detection (AD), used in computer-aided diagnosis (CAD), is more challenging due to sparse data, privacy concerns, and complex imagery [15, 53]. While ML algorithms have been used to detect lung diseases from chest X-rays [5], others [45] note that foreign particle infiltration (a symptom of Tuberculosis) is harder to identify. This review compares state-of-the-art (SOTA) IAD methods, focusing on sample datasets from industrial applications (MVTec AD dataset [8]) and medical applications (NIH Chest X-ray dataset [54]). We identify opportunities to adapt industrial IAD advancements to improve medical IAD, particularly for detecting lung infiltration, potentially enhancing the diagnostic utility of common chest X-rays.