GRU-powered sleep stage classification with permutation-based EEG channel selection.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Accurate sleep stage classification is crucial for diagnosing sleep disorders.
• Traditional methods often rely on high-density electroencephalographic (EEG) data, which can be computationally expensive.
• Deep learning models show promise but require efficient feature selection strategies.

Purpose of the Study:

• To develop a computationally inexpensive permutation-based method for selecting informative electroencephalographic (EEG) channels for sleep stage classification.
• To evaluate the effectiveness of this method using a gated recurrent unit (GRU) deep learning model.
• To compare the performance of permutation-selected channels against standard recommendations.

Main Methods:

• Systematic permutation of electroencephalographic (EEG) channels to identify optimal subsets for 5-class sleep stage classification.
• Utilizing a gated recurrent unit (GRU) deep learning model for analyzing EEG data.
• Analysis of an EEG dataset from the International Institute for Integrative Sleep Medicine (WPI-IIIS).

Main Results:

• Performance significantly decreases with fewer than 3 EEG channels.
• 3 randomly selected channels via permutation match or exceed the predictive accuracy of 3 channels recommended by the American Academy of Sleep Medicine (AASM).
• The N1 sleep stage shows the largest drop in prediction accuracy when channel density is reduced.

Conclusions:

• Permutation-based channel selection effectively identifies informative EEG channels, maintaining or enhancing model efficiency.
• The GRU model demonstrates strong capability in retaining temporal information for accurate sleep stage prediction.
• This approach offers a computationally efficient alternative to high-density EEG for sleep stage classification.