Back to all papers

Artificial intelligence-generated color overlay for ultrasound identification of the brachial plexus in regional anesthesia: an external validation study.

July 21, 2026pubmed logopapers

Authors

Kamimura Y,Aota Y,Oba S,Tomonari T,Kawatsu A,Sobue K

Affiliations (2)

  • Department of Anesthesiology and Intensive Care Medicine, Nagoya City University Graduate School of Medical Sciences, 1 Kawasumi, Mizuho-cho, Mizuho-ku, Nagoya, 467-8601, Japan. [email protected].
  • Department of Anesthesiology and Intensive Care Medicine, Nagoya City University Graduate School of Medical Sciences, 1 Kawasumi, Mizuho-cho, Mizuho-ku, Nagoya, 467-8601, Japan.

Abstract

Accurate identification of neural structures is essential for safe ultrasound-guided regional anesthesia. Although artificial intelligence (AI)-generated color overlay has been introduced to support anatomical recognition, external validation using clinically oriented evaluation methods remains limited. We evaluated the diagnostic accuracy of an AI-generated color overlay for visualizing neural structures corresponding to the C5 and C6 nerve roots. In this prospective observational study, two anesthesiology trainees performed bilateral interscalene ultrasound scans in 20 healthy adult volunteers. During live scanning, ultrasound-system video recordings were saved from the initiation of scanning to acquisition of an optimal image, and still images were captured at the optimal image. Subsequently, with the operators blinded to the AI output, the color overlay generated in real time during ultrasound scanning was evaluated. Two board-certified anesthesiologists independently evaluated the datasets and assessed the accuracy of the color overlay in visualizing neural structures corresponding to the C5 and C6 nerve roots using a four-category classification: true positive, true negative, false positive, and false negative. Of the 160 evaluated nerve structures, six structures without expert consensus were excluded from analysis. Overall sensitivity, specificity, and accuracy were 93.8% (95% confidence interval [CI], 88.6-96.7), 88.9% (95% CI, 56.5-98.0), and 93.5% (95% CI, 88.5-96.4), respectively. Among all evaluated structures, false-positive and false-negative findings accounted for 0.7% and 5.8%, respectively. AI-generated color overlay demonstrated high accuracy in visualizing neural structures corresponding to the C5 and C6 nerve roots at the interscalene level on ultrasound images. This technology may enhance anatomical recognition and serve as a useful diagnostic and educational support tool, while complementing, rather than replacing, clinicians' own interpretation.

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.