An Attention-Based Deep Learning Approach for Sleep Stage Classification With Single-Channel EEG.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Neuroscience

Background:

• Accurate sleep stage classification is crucial for assessing sleep quality.
• Current methods for sleep analysis often require complex multichannel recordings.

Purpose of the Study:

• To propose a novel attention-based deep learning architecture, AttnSleep, for classifying sleep stages.
• To utilize single-channel electroencephalogram (EEG) signals for improved sleep analysis.

Main Methods:

• Developed AttnSleep, featuring a multi-resolution convolutional neural network (MRCNN) and adaptive feature recalibration (AFR) for feature extraction.
• Employed a temporal context encoder (TCE) with multi-head attention and causal convolutions to capture temporal dependencies.
• Evaluated the model on three public sleep datasets.

Main Results:

• AttnSleep demonstrated superior performance compared to state-of-the-art techniques.
• The model achieved high accuracy across various evaluation metrics on public datasets.
• The proposed architecture effectively extracts and models temporal features from single-channel EEG.

Conclusions:

• AttnSleep offers a powerful and efficient approach for automatic sleep stage classification.
• The model's performance highlights the potential of attention-based deep learning for sleep analysis using single-channel EEG.
• The study provides open-source code and data for reproducibility.