EEG gamma frequency and sleep-wake scoring in mice: comparing two types of supervised classifiers

Area of Science:

• Neuroscience
• Sleep Research
• Animal Models

Background:

• Growing demand for circadian rhythm screening in genetically modified animals necessitates reliable sleep stage scoring.
• Existing methods lack adaptability and consistent variable selection for accurate sleep scoring.
• Electroencephalography (EEG) and electromyography (EMG) are key for monitoring sleep-wake states.

Purpose of the Study:

• To develop and validate a reliable automated sleep stage scoring program for freely moving mice.
• To identify optimal frequency variables and classifiers for distinguishing sleep stages.
• To compare the performance of Linear Discriminant Analysis (LDA) with other methods.

Main Methods:

• EEG and neck muscle EMG were recorded from freely moving C57BL/6 mice.
• Analysis included conventional power spectral density and period-amplitude analysis of frequency variables.
• Manual staging was compared against supervised classifiers: LDA and Classification Tree.

Main Results:

• Four variables, including gamma, delta, and upper theta band amplitudes and neck muscle EMG, were most effective for sleep-wake separation.
• LDA outperformed Classification Tree and a conventional threshold formula, especially with limited training data.
• Optimal epoch duration for analysis was found to be 8 to 10 seconds.

Conclusions:

• Gamma and upper theta activity are particularly indicative of REM sleep and waking states.
• LDA provides a robust and efficient method for supervised, semi-automatic sleep scoring.
• This approach significantly reduces the time and effort required for sleep analysis in animal research.