OSA severity assessment based on sleep breathing analysis using ambient microphone.

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Obstructive sleep apnea (OSA) is a common sleep disorder.
• Accurate OSA severity estimation is crucial for effective treatment.
• Current diagnostic methods like polysomnography (PSG) can be cumbersome.

Purpose of the Study:

• To develop and validate an audio-based system for estimating obstructive sleep apnea (OSA) severity.
• To assess the system's ability to predict the apnea-hypopnea index (AHI).
• To establish a reliable, non-contact alternative for OSA severity assessment.

Main Methods:

• A Gaussian mixture regression algorithm was employed for AHI estimation.
• Feature selection using a genetic algorithm identified optimal time and spectral domain features.
• The system was trained and validated on full-night audio recordings from 155 subjects undergoing PSG.

Main Results:

• The audio-based system achieved a high Pearson correlation (R=0.89) with PSG AHI scores.
• The system demonstrated a low AHI error of 7.35 events/hr.
• Diagnostic agreement with PSG reached 77.3% for OSA severity classification.

Conclusions:

• The proposed audio-based system shows promising performance for non-contact OSA severity estimation.
• The system provides a reliable alternative to traditional methods for AHI assessment.
• Further research can explore integration into wearable devices for remote OSA monitoring.