Authors: Nakamura H, et al.
Journal of Clinical Anesthesia, 2025. doi:10.1016/j.jclinane.2025.111914
This study developed an AI-assisted model based on YOLOv8n to identify laryngeal structures in infants and neonates during video laryngoscopy. Using 1,197 images for training/validation and 399 images for testing, the model demonstrated strong performance with a sensitivity of 0.74, specificity of 0.99, and an AUC of 0.91. It successfully identified the opening vocal cords and arytenoids, even when only the arytenoids were visible. However, errors included esophageal misidentification and difficulty detecting the larynx when obstructed.
The results suggest that AI can assist clinicians in improving visualization of the airway during neonatal and infant intubation, but refinement is still required before routine clinical adoption.
Practical takeaway: AI-assisted video laryngoscopy may become a valuable adjunct for challenging neonatal and infant intubations, but current limitations—particularly misidentification risks—mean clinicians must use AI cautiously and not replace expert judgment.
Thank you to the Journal of Clinical Anesthesia for publishing this work.