EOGNET: A Novel Deep Learning Model for Sleep Stage Classification Based on Single-Channel EOG Signal.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Automatic sleep staging is crucial for diagnosing sleep disorders.
• Electroencephalography (EEG) is the standard but is cumbersome for continuous monitoring.
• Electrooculogram (EOG) signals offer a more convenient alternative for sleep data acquisition.

Purpose of the Study:

• To develop and validate a novel sleep staging method utilizing electrooculogram (EOG) signals.
• To demonstrate the feasibility of EOG-based sleep staging as a practical alternative to EEG.
• To compare the performance of the proposed EOG method against established EEG-based approaches and state-of-the-art techniques.

Main Methods:

• A hybrid deep learning model combining a two-scale convolutional neural network (CNN) and a recurrent neural network (RNN).
• The CNN extracts epoch-wise features from raw EOG signals, while the RNN captures temporal dependencies.
• Validation was performed on 101 full-night sleep datasets from the Montreal Archive of Sleep Studies and Sleep-EDF databases.

Main Results:

• The proposed EOG-based method achieved an overall accuracy of 81.2% on the Montreal Archive of Sleep Studies dataset.
• An accuracy of 76.3% was obtained on the Sleep-EDF dataset.
• Performance was comparable to existing methods that utilize EEG signals, outperforming six state-of-the-art approaches.

Conclusions:

• The developed EOG-based sleep staging method is effective and practical for sleep monitoring.
• This approach offers a convenient and accurate alternative to traditional EEG-based sleep analysis.
• The study opens new possibilities for non-invasive, long-term sleep monitoring solutions.