SegWay: A simple framework for unsupervised sleep segmentation in experimental EEG recordings.

Area of Science:

• Neuroscience
• Animal Behavior
• Sleep Science

Background:

• Sleep analysis in animal models commonly uses electroencephalogram (EEG) and electromyogram (EMG) recordings.
• Manual or algorithmic scoring of vigilance states (Wake, REM, NREM) is standard but presents reproducibility and accuracy challenges.
• Existing computer algorithms often require extensive user input or complex computational steps, limiting accessibility for researchers.

Purpose of the Study:

• To develop and validate an unsupervised sleep segmentation framework, named SegWay, for analyzing animal sleep data.
• To enhance the accuracy of sleep scoring and the estimation of key sleep metrics by incorporating state transition modeling.
• To provide a more accessible and reproducible method for sleep analysis in animal models.

Main Methods:

• An unsupervised sleep segmentation framework, SegWay, was developed.
• The algorithm was applied step-by-step to unlabeled EEG recordings from mice.
• Sleep scoring accuracy and sleep metrics estimation were validated against manual scoring by researchers.

Main Results:

• SegWay demonstrated accurate sleep scoring and estimation of sleep metrics.
• Modeling transitions between vigilance states improved the accuracy of sleep scoring and metric calculation.
• The framework offers a more reproducible approach compared to traditional methods.

Conclusions:

• The SegWay framework provides an accurate and accessible method for unsupervised sleep analysis in animal models.
• Incorporating state transition modeling significantly enhances sleep scoring and metric estimation.
• SegWay offers a valuable tool for advancing sleep research in animal models by improving reproducibility and accuracy.