Deep learning‐based facial analysis for predicting difficult videolaryngoscopy: a feasibility study
Authors' conclusion
Artificial intelligence-based facial analysis is a feasible technique for predicting difficulty during videolaryngoscopy, and the model developed using neural networks has higher predictive performance than traditional methods.
Study context
- Population
- Patients undergoing videolaryngoscopy for tracheal intubation514 difficult videolaryngoscopy
- Participants
- 5,849 patients
- Comparison
- AI facial model vs bedside examination and multivariate scores
- Outcome
- Difficult videolaryngoscopy (Cormack-Lehane grade 3 or 4)
Methodological detailsfeasibility study, deep learning (ResNet-18), machine learning prediction model
- Feasibility study
- Deep learning (ResNet-18)
- Machine learning prediction model
Performance
As reported in the abstract
Prediction of difficult videolaryngoscopy (test set)
| Test or model | AUC (95% CI 0.733-0.825) | Sensitivity (95% CI 0.650-0.845) | Specificity (95% CI 0.626-0.794) |
|---|---|---|---|
| Facial model | 0.779 | 0.757 | 0.721 |
Facial model
AUC (95% CI 0.733-0.825)0.779
Sensitivity (95% CI 0.650-0.845)0.757
Specificity (95% CI 0.626-0.794)0.721
- Facial model vs Bedside examination and multivariate scores (El-Ganzouri and Wilson): P < 0.001 — significantly higher with facial model
Original publication
Deep learning‐based facial analysis for predicting difficult videolaryngoscopy: a feasibility study. Anaesthesia. Published 13 December 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →