Variability between human experts and artificial intelligence in identification of anatomical structures by ultrasound in regional anaesthesia: a framework for evaluation of assistive artificial intelligence
Authors' conclusion
Human experts and the AI system both showed the same pattern of agreement in sono-anatomical structure identification.
Study context
- Population
- Experts in ultrasound-guided regional anaesthesia (UGRA) annotating 100 structures in 30 ultrasound videos across six anatomical regions30 videos · 100 structures
- Participants
- 19 experts
- Comparison
- Human-human vs Human-AI
- Human-humann = 19reference
- Outcome
- Agreement (Dice score) in identification of sono-anatomical structures, by structure class and structure
Main finding
As reported in the abstract
Mean Dice score by structure class
Arteries, human-human/human-AI
0.88/0.86
Muscles, human-human/human-AI
0.80/0.77
Nerves, human-human/human-AI
0.48/0.41
Further findings
Dice score by individual structure
- Highest: sartorius muscle, human-human/human-AI
- 0.91/0.92
- Lowest: radial nerve, human-human/human-AI
- 0.21/0.27
- Wide discrepancy in consistency for different structures
Original publication
Variability between human experts and artificial intelligence in identification of anatomical structures by ultrasound in regional anaesthesia: a framework for evaluation of assistive artificial intelligence. Br J Anaesth. Published 26 October 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →