Sleep-wake transition in narcolepsy and healthy controls using a support vector machine.

Area of Science:

• Neuroscience
• Sleep Medicine
• Biomedical Engineering

Background:

• Narcolepsy is a sleep disorder marked by disrupted sleep-wake cycles.
• Manual scoring of polysomnography (PSG) is the standard for identifying sleep-wake transitions.
• Automated methods are needed to improve the efficiency and objectivity of sleep-wake transition detection.

Purpose of the Study:

• To develop and validate an automatic classifier for distinguishing sleep and wake epochs using single-channel EEG.
• To assess the classifier's performance against manual scoring in narcolepsy patients and healthy controls.

Main Methods:

• Utilized wavelet packet transformation to extract features from specific EEG frequency bands.
• Employed a support vector machine (SVM) classifier for sleep-wake epoch classification.
• Validated the classifier using hold-out and 10-fold cross-validation on PSG data from 47 narcolepsy patients and 15 healthy controls.

Main Results:

• The automatic classifier achieved 90% accuracy in hold-out validation and 88% in 10-fold cross-validation.
• High performance metrics were observed, including Area Under the Receiver Operating Characteristic Curve (AUC) of 95% (hold-out) and 92% (cross-validation).
• Narcolepsy with cataplexy patients exhibited significantly more nocturnal sleep-wake transitions compared to narcolepsy without cataplexy and healthy subjects.

Conclusions:

• The developed automatic classifier demonstrates high validity for identifying sleep-wake transitions.
• Increased nocturnal sleep-wake transitions in narcolepsy with cataplexy suggest underlying instability in the sleep-wake regulatory system.
• This automated approach offers a promising tool for objective sleep-wake analysis in narcolepsy research and clinical practice.