Automatic sleep stage classification using ear-EEG.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Sleep Medicine

Background:

• Polysomnography is standard for sleep disorder diagnosis but is cumbersome and labor-intensive.
• Automatic sleep staging shows promise in overcoming manual assessment limitations.
• Ear-EEG presents a minimally invasive approach for electroencephalogram (EEG) recording.

Purpose of the Study:

• To evaluate the feasibility of automatic sleep stage classification using ear-EEG.
• To compare the performance of ear-EEG based classification with traditional scalp EEG methods.

Main Methods:

• A preliminary study involved 18 subjects, with sleep scoring by a clinical expert using polysomnography (EEG, EOG, EMG).
• A single-channel EEG sleep stage classifier, based on prior work, was applied to both scalp and ear-EEG data.
• Performance was assessed using 10-fold cross-validation (inter-subject for scalp EEG, intra-subject for ear-EEG).

Main Results:

• Scalp EEG classification achieved 85.7% agreement with expert scoring.
• Ear-EEG classification showed 82% agreement with expert scoring.
• Five subjects were excluded due to alpha wave contamination.

Conclusions:

• Automatic sleep stage classification using ear-EEG demonstrates comparable performance to single-channel scalp EEG.
• Ear-EEG is a potentially feasible technology for future minimally intrusive sleep stage classification.
• Further research is warranted to validate these findings in larger cohorts.