Deep Convolutional Recurrent Model for Automatic Scoring Sleep Stages Based on Single-Lead ECG Signal.

Area of Science:

• Artificial Intelligence
• Biomedical Engineering
• Sleep Medicine

Background:

• Manual sleep stage scoring is subjective and time-consuming.
• Accurate, automated sleep stage scoring is needed for quantitative sleep analysis.
• Electrocardiogram (ECG) signals offer a potential data source for automated sleep analysis.

Purpose of the Study:

• To develop and evaluate a deep convolutional recurrent (DCR) model for automatic sleep stage scoring.
• To assess the model's performance using raw single-lead ECG signals.
• To determine the efficacy of the DCR model for both five-class and three-class sleep staging.

Main Methods:

• A DCR model integrating deep convolutional and recurrent neural networks was constructed.
• The model architecture included three convolutional and two recurrent layers, optimized with dropout and batch normalization.
• The DCR model was trained and tested on single-lead ECG data from 112 subjects (control and sleep apnea groups).

Main Results:

• The DCR model achieved 74.2% accuracy for five-class sleep stage scoring (wake, N1, N2, N3, REM).
• The model reached 86.4% accuracy for three-class sleep stage scoring (wake, NREM, REM).

Conclusions:

• The DCR model demonstrates superior performance compared to previous methods.
• This AI-driven approach shows potential as an alternative tool for sleep monitoring.
• The DCR model could be valuable for sleep screening applications.