Sleep-wake stages classification using heart rate signals from pulse oximetry

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Obstructive sleep apnea/hypopnea syndrome (OSAHS) diagnosis relies heavily on the apnea/hypopnea index (AHI).
• Current pulse oximetry screening methods for OSAHS approximate AHI by counting desaturation events, often neglecting patient sleep stages.
• Accurate sleep stage determination is crucial for precise AHI calculation and OSAHS assessment.

Purpose of the Study:

• To develop and validate an automatic system for determining patient wakefulness or sleep state using only heart rate (HR) signals from pulse oximetry.
• To explore the potential of HR signal analysis for enhancing OSAHS screening and AHI estimation.
• To investigate the efficacy of various signal processing and machine learning techniques for sleep stage classification.

Main Methods:

• Extracted 70 features from heart rate signals using entropy, complexity, frequency, and time-scale domain analyses, alongside classical statistics.
• Reduced feature space dimensionality from 70 to 40 using support vector machine-based forward feature selection and random forest feature importance.
• Trained and tested algorithms on 5000 patients from the Sleep Heart Health Study database, employing 10-fold cross-validation.

Main Results:

• Achieved classification performance metrics including 85.2% accuracy, 88.3% specificity, 79.0% sensitivity, 67.0% positive predictive value, and 91.3% negative predictive value.
• Demonstrated the feasibility of using HR signals for reliable sleep stage determination.
• Validated the system's performance through rigorous testing on a large patient cohort.

Conclusions:

• The developed automatic system effectively determines sleep stages using HR signals from pulse oximetry.
• This approach holds significant promise for improving the accuracy of AHI estimation in OSAHS screening.
• Integrating sleep stage detection into pulse oximetry could lead to more precise and accessible OSAHS diagnostics.