Sleep Stage Detection Using Tracheal Breathing Sounds: A Pilot Study

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Sleep stage detection is crucial for clinical assessments and sleep studies.
• Current methods rely on electroencephalogram (EEG) and electrooculogram (EOG) signals.
• There is a need for alternative, non-invasive sleep stage identification techniques.

Purpose of the Study:

• To investigate the feasibility of using tracheal breathing sounds for sleep stage detection.
• To determine if breathing sound patterns change across different sleep stages (Wakefulness, REM, Stage II).
• To explore the temporal stability of breathing sound patterns during sleep.

Main Methods:

• Recorded tracheal breathing sounds from 12 individuals undergoing polysomnography (PSG).
• Analyzed breathing sounds using higher-order statistical analysis, specifically the Hurst exponent from bispectra.
• Selected noise-free breathing cycles from wakefulness, REM, and Stage II sleep at different sleep periods.

Main Results:

• Distinct, non-overlapping clusters were identified for wakefulness, REM, and Stage II sleep based on breathing sounds.
• A simple linear classifier achieved 100% accuracy in distinguishing REM and Stage II sleep for each subject.
• Consistent patterns were observed for REM and Stage II sleep when segments were within 3 hours of each other.

Conclusions:

• Tracheal breathing sounds show potential as a non-invasive method for sleep stage detection.
• Breathing sound analysis, particularly using the Hurst exponent, can differentiate key sleep stages.
• This technique could serve as a valuable alternative or complementary method to traditional PSG in sleep research and diagnostics.