Towards an automatic narcolepsy detection on ambiguous sleep staging and sleep transition dynamics joint model

Area of Science:

• Neuroscience
• Sleep Medicine
• Biomedical Engineering

Background:

• Narcolepsy diagnosis is challenging due to fragmented sleep stages.
• Current automatic sleep staging methods for narcolepsy lack patient identification capabilities.
• Multiple Sleep Latency Test (MSLT) analysis can be improved for narcolepsy detection.

Purpose of the Study:

• To develop an automated narcolepsy detection method utilizing MSLT recordings.
• To integrate automatic sleep staging with sleep transition dynamics for improved diagnostic accuracy.
• To establish a computational framework for efficient narcolepsy identification from polysomnography data.

Main Methods:

• A two-phase model was developed using electroencephalogram (EEG) and electrooculogram (EOG) signals from MSLT.
• Phase 1 employed an EasyEnsemble classifier for automatic sleep staging.
• Phase 2 utilized Principal Component Analysis (PCA) and logistic regression to analyze sleep transition dynamics and output narcolepsy likelihood.

Main Results:

• The model achieved 87.5% accuracy, 80.0% sensitivity, and 92.9% specificity for narcolepsy detection.
• Automatic sleep staging accuracy was reduced in narcolepsy patients compared to non-narcoleptic individuals.
• The framework was validated on 24 participants, including 10 narcolepsy patients.

Conclusions:

• The proposed method offers an automated and efficient approach for distinguishing narcolepsy patients based on MSLT.
• This study highlights the potential of combining automatic sleep staging and sleep transition dynamics for narcolepsy diagnosis.
• The findings can aid clinicians and neurophysiologists in objective interpretation and diagnosis of narcolepsy.