Back to all papers

Deep learning algorithm for automatic detection of acute ischemic stroke on noncontrast brain CT.

May 31, 2026pubmed logopapers

Authors

Yun TJ,Choi JW,Na H,Han M,Jung WS,Lee JY,Yoo RE,Hwang IP,Choi SH

Affiliations (7)

  • Department of Radiology, Seoul National University Hospital, Seoul, Republic of Korea.
  • Institute of Radiation Medicine, Seoul National University Medical Research Center, Seoul, Republic of Korea.
  • Department of Radiology, Seoul National University College of Medicine, Seoul, Republic of Korea.
  • Department of Radiology, Ajou University School of Medicine, San 5, Wonchon-dong, Yeongtong-gu, Suwon, Republic of Korea. [email protected].
  • SK Inc. C&C, Sungnam, Republic of Korea.
  • Purple AI Inc., Seoul, Republic of Korea.
  • Department of Radiology, Ajou University School of Medicine, San 5, Wonchon-dong, Yeongtong-gu, Suwon, Republic of Korea.

Abstract

This study aimed to evaluate the diagnostic performance of a deep learning-based algorithm for detecting acute ischemic stroke (AIS), including small infarcts, on non-contrast computed tomography (NCCT). A retrospective, multi-reader, pivotal, crossover, randomized study involving 917 cases was conducted to validate the performance of an artificial intelligence (AI) algorithm. NCCT images were independently reviewed by nine readers categorized into three subgroups (three non-radiologist physicians, three board-certified radiologists, and three neuroradiologists), both with and without the assistance of the AI algorithm. In standalone analysis, the AI model achieved an area under the receiver operating characteristic curve of 0.8144, with an accuracy of 73.8%, sensitivity of 75.8%, and specificity of 72.6%. AI-assisted brain CT interpretation showed higher diagnostic accuracy than AI-unassisted interpretation (75.63% vs. 72.03%, with improvement of 3.60%, p < 0.001). Among the three subgroups of reviewers, non-radiologist physicians showed the greatest improvement in diagnostic accuracy with AI assistance compared with without AI assistance (75.35% vs. 69.97%, with improvement of 5.38%, p < 0.001). Our results suggested that AI-assisted interpretation can improve diagnostic performance in the detection of AIS on NCCT across a range of infarct volumes and reader expertise levels. Its integration into clinical workflows could facilitate more timely and equitable AIS diagnosis.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.