Characterizing transition state in mouse vigilance with electroencephalogram-electromyogram hypnodensity

Area of Science:

• Neuroscience
• Computational Biology
• Sleep Research

Background:

• Conventional rodent sleep scoring uses discrete epochs, obscuring continuous vigilance-state transitions.
• A standardized framework for characterizing intermediate sleep-wake states in rodents is lacking.
• Understanding these transitions is crucial for studying arousal stability and neurological disorders.

Purpose of the Study:

• To characterize the temporal dynamics of vigilance-state transitions in mice.
• To validate a machine-learning approach for objective detection of these transitions.
• To provide a reproducible method for quantifying sleep-wake dynamics.

Main Methods:

• Chronic EEG/EMG recordings were obtained from C57BL/6N mice.
• Expert annotators manually scored transition onset and duration.
• Support Vector Machine (SVM) classifiers were trained using quantitative EEG/EMG features.

Main Results:

• Inter-rater reliability among experts was moderate to low, highlighting scoring ambiguity.
• Non-Rapid Eye Movement Sleep-to-Rapid Eye Movement Sleep (NREMS-to-REMS) transitions were longest; REMS-to-NREMS were most abrupt.
• SVM models accurately predicted transition midpoints despite human scoring variability.

Conclusions:

• A machine-learning approach effectively quantifies rodent sleep-wake transitions.
• This method offers a reproducible framework for studying arousal stability.
• The approach is vital for investigating sleep impairments in disease models.