A novel expert classifier approach to pre-screening obstructive sleep apnea during wakefulness

Area of Science:

• Medical Diagnostics
• Biomedical Engineering
• Pulmonology

Background:

• Obstructive sleep apnea (OSA) diagnosis relies on polysomnography (PSG), which is time-consuming and inconvenient.
• There is a need for accessible, cost-effective methods to pre-screen for severe OSA.

Purpose of the Study:

• To further develop and validate the Awake-OSA method for predicting severe OSA using wakeful breathing sounds.
• To assess the efficacy of an expert classification approach for OSA screening.

Main Methods:

• Utilized breath sound features from 249 subjects to train classifiers.
• Employed an expert classification approach with majority-voting classifiers trained to distinguish between non-OSA and severe-OSA classes.
• Combined classifier outcomes using a truth matrix for final classification of 180 subjects.

Main Results:

• Classified 79% of OSA and 75% of non-OSA subjects.
• Achieved a testing sensitivity of 78% and specificity of 86% for OSA classification.
• Demonstrated robust and generalizable performance through consistent training and testing accuracies.

Conclusions:

• The Awake-OSA method shows promise as a rapid, low-cost pre-screening tool for severe OSA in clinical settings.
• This technique can help physicians identify patients who require overnight PSG, optimizing diagnostic workflows.
• The expert classification approach provides reliable predictions based on readily available breath sound data.