Objective Pharyngeal Phenotyping in Obstructive Sleep Apnea With High-Resolution Manometry.

Area of Science:

• Sleep Medicine
• Medical Engineering
• Otolaryngology

Background:

• Obstructive sleep apnea (OSA) diagnosis often relies on drug-induced sleep endoscopy (DISE).
• DISE is costly, subjective, and requires sedation.
• Objective, dynamic upper airway phenotyping is needed for OSA surgical selection.

Purpose of the Study:

• To investigate high-resolution pharyngeal manometry (HRM) for upper airway phenotyping in OSA.
• To develop a software system using HRM to predict pharyngeal collapse sites.
• To compare HRM-based predictions with DISE findings.

Main Methods:

• Prospective cross-sectional study involving 40 participants.
• Simultaneous HRM and DISE recordings were obtained.
• A machine learning algorithm was developed to predict collapse severity from HRM data, validated against expert DISE review.

Main Results:

• The machine learning model achieved high performance during training (average F1-score 0.86, accuracy 0.91).
• A K-nearest neighbor model on a holdout set demonstrated excellent accuracy (F1-score 0.96, accuracy 0.97).
• Prediction of complete concentric palatal collapse yielded an F1-score of 0.86.

Conclusions:

• HRM shows potential for objective and dynamic pharyngeal mapping in OSA.
• This technique may provide a reliable and reproducible method for assessing upper airway anatomy.
• HRM could offer a new diagnostic pathway for OSA patients requiring surgical intervention.