Automatic sleep staging by cardiorespiratory signals: a systematic review.

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Electroencephalography (EEG) signal recording and analysis present challenges.
• Cardiorespiratory signals offer a viable alternative for automatic sleep staging.
• Improving automatic sleep staging using cardiorespiratory signals is crucial for sleep research.

Purpose of the Study:

• To identify critical factors for enhancing automatic sleep staging accuracy using cardiorespiratory signals.
• To provide recommendations for optimizing cardiorespiratory-based sleep staging.

Main Methods:

• A systematic literature review was conducted.
• Analysis focused on feature extraction epoch length, signal time delays, ECG morphology, and deep learning architectures.
• Key findings were synthesized to propose improvements.

Main Results:

• A 4.5-minute epoch length is recommended over the standard 30-second segments for feature extraction.
• Accounting for the time delay between central and autonomic nervous system signals is important.
• Electrocardiogram (ECG) signal morphology provides valuable information for staging.
• Simultaneous application of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks on large polysomnography (PSG) databases enhances reliability.

Conclusions:

• Integrating optimal epoch length, time delay considerations, ECG morphology, and advanced deep learning models (CNN, LSTM) can significantly improve cardiorespiratory-based sleep staging.
• This approach aims to achieve results comparable to EEG-based staging, overcoming EEG's technical limitations.