Voice Analysis as a Method for Preoperatively Predicting a Difficult Airway Based on Machine Learning Algorithms:...

Area of Science:

• Anesthesiology and airway management.
• Machine learning applications in healthcare.
• Acoustic analysis and signal processing.

Background:

• Difficult airway management poses significant risks for anesthesiologists.
• Current anthropometric screening methods for airway difficulty have limited predictive value.
• Novel approaches are needed for accurate preoperative airway assessment.

Purpose of the Study:

• To investigate the efficacy of voice analysis using machine learning algorithms for predicting difficult airways.
• To compare the performance of voice-based models against traditional airway assessment techniques.

Main Methods:

• An observational, multicenter study included 313 adult patients undergoing general anesthesia with endotracheal intubation.
• Voice recordings (vowels "A, E, I, O, U") were captured in various head positions.
• Machine learning models were developed using demographic data and acoustic voice parameters (e.g., Shimmer, Jitter, HNR).

Main Results:

• Two machine learning models demonstrated high performance in predicting difficult airways (AUC 0.91 and 0.90).
• Models focusing on Cormack grades I and IV showed the most distinct predictive capabilities.
• Demographic data combined with specific vowel sounds and voice parameters yielded the best results.

Conclusions:

• Machine learning analysis of acoustic voice parameters, alongside demographic data, shows significant potential for predicting difficult airways.
• This approach offers a promising, non-invasive tool for enhancing preoperative airway assessment.
• Further research can refine these models for clinical application to improve patient safety.