Automatic Sleep Staging Using Cardiorespiratory Signals: A Systematic Review of Methodologies and Performance

Area of Science:

• Cardiology
• Sleep Medicine
• Biomedical Engineering

Background:

• Polysomnography (PSG) is the traditional gold standard for sleep studies but is resource-intensive.
• Cardiorespiratory signals present a scalable alternative for longitudinal sleep monitoring and health management.
• Automatic sleep staging using cardiorespiratory data is an emerging field with significant potential.

Purpose of the Study:

• To systematically review the methodologies and performance of automatic sleep staging using cardiorespiratory signals.
• To identify the accuracy, limitations, and potential for clinical translation of these methods.
• To provide recommendations for future research and development.

Main Methods:

• Systematic literature search across four databases for studies published since 2010.
• Analysis of 35 identified studies focusing on cardiorespiratory signal-based sleep staging.
• Comparative assessment of signal modalities and machine learning algorithms.

Main Results:

• Cardiorespiratory signal-based sleep staging achieved a mean accuracy of 70%.
• No significant performance differences were found between different signal types or between traditional machine learning and deep learning algorithms.
• Significant heterogeneity in methodology, lack of external validation, poor N1 stage classification, and limited generalization were identified as critical limitations.

Conclusions:

• Cardiorespiratory-based sleep staging shows practical accuracy but faces challenges in clinical translation.
• Standardized guidelines and validation on diverse datasets are crucial for future development.
• This technology can potentially support population-level screening and clinical decision-making if advancements are made.