Real-time AI-assisted nerve tracking for ultrasound-guided nerve identification: a prospective randomized trial
In brief
AI-assisted training does not raise novice nerve ID success from 58% to 66%
In a randomized trial of 78 ultrasound novices, first-attempt identification of the median nerve and supraclavicular brachial plexus was similar whether participants used real-time AI tracking or standard B-mode training. Additional AI-guided practice improved confidence modestly but did not translate into higher independent success, leaving its clinical value uncertain.
- Journal
- Regional anesthesia and pain medicine (Q1)
- Published
- 18 September 2026
- Study design
- Randomized controlled trial
- Evidence level
- Level 1, High (CEBM 1b)
- Authors
- Jeongsoo Kim, Seung Pyo Nam, Jehyuk Yoo, Yoontae Hong, Hungjun Joo, Yongjae Yoo, et al.
- PMID
- 42760064
- DOI
- 10.1136/rapm-2026-108180
Why clinicians should know about it
- Picked for Anatomy (top studies of the week, 20 September 2026): Real‑time AI‑assisted nerve tracking for ultrasound‑guided nerve identification
Abstract
BACKGROUND: Artificial intelligence (AI)-assisted nerve tracking facilitates real-time ultrasound-guided nerve identification, but its effect on independent nerve identification after AI support is removed remains uncertain. METHODS: In this prospective randomized trial, 80 ultrasound novices (60 medical students and 20 anesthesiology residents) were assigned to either a control group receiving B-mode ultrasound training without AI assistance or an AI-assisted group using a real-time nerve tracking system (NerveTrack). Following a standardized lecture and a 10-minute hands-on training session, participants underwent assessment with the AI function disabled. The primary outcome was first-attempt independent identification success for the median nerve and supraclavicular brachial plexus. Secondary outcomes included performance after additional training and self-reported confidence. RESULTS: 78 participants completed the study. First-attempt independent identification success rates did not differ significantly between the control and AI-assisted groups for the median nerve (57.5% vs 65.8%; RR 1.14, 95% CI 0.81 to 1.62; p=0.452) or the brachial plexus (35.0% vs 42.1%; RR 1.20, 95% CI 0.68 to 2.11; p=0.519). Following additional AI-assisted training, participants in the AI-assisted group reported greater confidence in identifying the brachial plexus than those in the control group (4.7±0.5 vs 4.4±0.9; p=0.046). No significant differences were observed in independent identification success after additional training. CONCLUSIONS: Brief AI-assisted training did not improve immediate independent ultrasound identification of the median nerve or supraclavicular brachial plexus after AI support was removed. These findings do not establish long-term retention, transfer to unfamiliar anatomy, or clinical procedural performance. THE REGISTRATION NUMBER: NCT05857787.
Abstract as published, via PubMed.
For healthcare professionals. The summary is generated by AI from the published abstract, and the evidence level is assigned automatically from the study design on the Oxford CEBM hierarchy. Neither is medical advice. Read the full paper before changing practice.