2.9. EEG/EMG scoring

We scored all recording files using the online computational tool SPINDLE (Miladinović et al., ²⁰¹⁹) for animal sleep data (https://sleeplearning.ethz.ch/). In short, European Data Format (.edf) files, consisting of two parietal EEG and one nuchal EMG channels, were uploaded to SPINDLE to retrieved vigilance states with 4‐s epoch resolution. The algorithm classified three vigilance states: wakefulness; NREM sleep and REM sleep. Wakefulness was defined based on high or phasic EMG activity for more than 50% of the epoch duration, and low‐amplitude but high‐frequency EEG. NREM sleep was characterised by reduced or no EMG activity, increased EEG power in the frequency band < 4 Hz, and the presence of slow oscillations. REM sleep was defined based on high theta power (6–9 Hz frequency band) and low muscle tone. Additionally, unclear epochs containing data outliers or signal perturbations related to environmental interference rather than changes in brain state were labelled as artefacts in wakefulness, NREM or REM sleep.