Automatic sleep stage classification based on EEG signals by using neural networks and wavelet packet coefficients

Area of Science:

• Biomedical Engineering
• Neuroscience
• Sleep Medicine

Background:

• Sleep disorders affect a significant global population, necessitating efficient diagnostic tools.
• Current sleep stage classification relies on time-consuming manual analysis of biomedical signals like EEG.
• Distinguishing between certain sleep stages (e.g., NREM Stage 1 and REM sleep) using EEG alone presents a challenge.

Purpose of the Study:

• To develop and evaluate an automated method for classifying four distinct sleep stages using only EEG signals.
• To address the similarity between NREM Stage 1 and REM sleep by grouping them for classification.
• To provide a more efficient alternative to manual sleep scoring.

Main Methods:

• Utilized electroencephalography (EEG) signals for sleep stage classification.
• Employed wavelet packet coefficients for feature extraction from EEG data.
• Applied artificial neural networks (ANNs) for automated classification of sleep stages.
• Used seven all-night polysomnography recordings from the Physionet database.

Main Results:

• Successfully discriminated between four sleep stages: Awake, Stage 1 + REM, Stage 2, and Slow Wave Stage.
• Achieved a specificity of 94.4 +/- 4.5%.
• Obtained a sensitivity of 84.2 +/- 3.9%.
• Reached an overall accuracy of 93.0 +/- 4.0% in automated sleep stage classification.

Conclusions:

• Automated classification of key sleep stages using EEG signals is feasible and accurate.
• The proposed method, combining wavelet packet coefficients and ANNs, offers a promising tool for sleep analysis.
• This automated approach can significantly reduce the time and effort required for sleep disorder diagnosis.