Automatic classification of sleep stages using EEG signals and convolutional neural networks

Area of Science:

• Neuroscience
• Medical Technology
• Artificial Intelligence

Background:

• Sleep stage classification is vital for assessing lifestyle quality and diagnosing diseases like diabetes and obesity.
• Electroencephalography (EEG) signals are commonly used for sleep staging analysis.
• Abnormal sleep patterns are linked to various health issues.

Purpose of the Study:

• To propose a robust methodology for sleep stage classification using EEG signals.
• To leverage a 2D convolutional neural network (CNN) for enhanced sleep staging accuracy.
• To evaluate the performance of the CNN-based approach across multiple EEG channels.

Main Methods:

• EEG signals were segmented into 30-second epochs.
• Epochs were converted into 2D time-frequency analysis images.
• A 2D CNN was employed to classify sleep stages based on these images.

Main Results:

• The proposed CNN methodology achieved a high accuracy of 99.39% for the C4-A1 channel.
• All other channels demonstrated accuracy above 98.5%.
• The methodology outperformed existing literature in terms of accuracy.

Conclusions:

• The developed CNN-based method is robust and highly accurate for sleep stage classification.
• Individual EEG channels can be effectively utilized for precise sleep staging.
• This approach offers a valuable tool for physicians, particularly neurologists, in diagnosing sleep-related diseases.