Automatic Sleep Stage Classification Using Single-Channel EEG: Learning Sequential Features with Attention-Based...

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Machine Learning

Background:

• Accurate sleep stage classification is crucial for diagnosing sleep disorders.
• Traditional methods often rely on manual feature extraction, which can be time-consuming and subjective.
• Automated sleep stage classification using electroencephalogram (EEG) signals offers a promising alternative.

Purpose of the Study:

• To develop and evaluate a novel feature learning approach for single-channel automatic sleep stage classification.
• To leverage deep bidirectional recurrent neural networks (RNNs) with an attention mechanism for enhanced classification performance.
• To investigate the impact of a learned filter bank on preprocessing EEG feature vectors.

Main Methods:

• Decomposition of EEG epochs into smaller frames and transformation into frame-wise feature vectors.
• Training an attention-based RNN in a sequence-to-label manner for sleep stage classification.
• Utilizing the trained network as a feature extractor, followed by classification with a linear Support Vector Machine (SVM).
• Proposing a discriminative method to learn a filter bank using a Deep Neural Network (DNN) for preprocessing.

Main Results:

• The attention-based RNN effectively encodes sequence information into high-level feature vectors.
• The proposed feature learning approach demonstrated good performance on the Sleep-EDF dataset.
• Preprocessing frame-wise feature vectors with the learned filter bank led to further improvements in classification accuracy.

Conclusions:

• Deep bidirectional RNNs with attention mechanisms provide a powerful framework for automatic sleep stage classification.
• The proposed method offers an effective strategy for learning discriminative features directly from EEG data.
• The integration of a learned filter bank enhances the robustness and performance of the sleep stage classification system.