Automatic and Accurate Sleep Stage Classification via a Convolutional Deep Neural Network and Nanomembrane Electrodes

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Sleep stage classification is crucial for diagnosing sleep disorders.
• Manual scoring is prone to human error and inconsistency.
• Automatic methods, particularly machine learning, offer higher efficiency.

Purpose of the Study:

• To develop a highly accurate, automatic sleep stage classification method using a convolutional neural network (CNN).
• To validate the CNN model's reliability on standard polysomnography data and its transferability to data from wearable electrodes.

Main Methods:

• A CNN model was trained and validated using a public dataset with electroencephalogram (EEG) and electrooculogram (EOG) signals.
• The model incorporated multi-taper spectrogram pre-processing.
• Validation was performed on a new dataset acquired using wearable nanomembrane dry electrodes.

Main Results:

• The CNN model achieved 88.85% training accuracy on the validation dataset.
• The model demonstrated 81.52% prediction accuracy on the laboratory dataset measured with wearable electrodes.
• Results confirm the method's reliability and transferability.

Conclusions:

• The developed CNN-based method provides a reliable and accurate approach for automatic sleep stage classification.
• The model shows strong performance and transferability, applicable to data from both standard polysomnography and novel wearable sensors.
• This advancement can enhance the efficiency and consistency of sleep disorder diagnosis.