Probabilistic sleep staging in MSLTs across hypersomnia disorders

Area of Science:

• Sleep Medicine
• Neurology
• Biomarker Discovery

Background:

• Narcolepsy type 1 (NT1) is a chronic neurological disorder characterized by excessive daytime sleepiness.
• Current diagnostic criteria for narcolepsy rely on Multiple Sleep Latency Test (MSLT) findings, but distinguishing between subtypes can be challenging.
• Investigating novel features within MSLT recordings may improve diagnostic accuracy and understanding of NT1 pathophysiology.

Purpose of the Study:

• To identify novel objective markers for narcolepsy type 1 (NT1) using features from both "lights on" (wake) and "lights off" (sleep) periods of the MSLT.
• To explore the potential of these MSLT features to differentiate NT1 from narcolepsy type 2 (NT2), idiopathic hypersomnia (IH), and subjective hypersomnia (sH).
• To determine if MSLTs from NT1 and NT2 patients can be distinguished despite adhering to the same diagnostic thresholds.

Main Methods:

• Analysis of 163 features from "lights on" and "lights off" periods of MSLT recordings in 177 patients with NT1, NT2, IH, and sH.
• Utilized automated probabilistic sleep staging (U-Sleep) to extract hypnodensities, quantifying sleepiness, microsleep, and sleep stage mixing.
• Employed machine learning approaches to differentiate hypersomnia subtypes based on extracted MSLT features.

Main Results:

• Features from "lights on" periods alone distinguished NT1 from NT2, IH, and sH with 76% sensitivity and 71% specificity.
• Incorporating features from all MSLT periods (lights on and off) improved the distinction between NT1 and NT2, achieving 77% sensitivity and 84% specificity.
• Analysis revealed that "lights off" features were particularly effective in distinguishing NT1 from NT2.

Conclusions:

• Microsleeps and sleep stage mixing during MSLT are significant indicators of the unstable sleep-wake states characteristic of NT1.
• "Lights off" MSLT features show promise in frequently distinguishing NT1 from NT2.
• These findings suggest that detailed analysis of MSLT recordings, including wake periods, can enhance the diagnostic accuracy for narcolepsy subtypes.