Sleep Identification Enabled by Supervised Training Algorithms (SIESTA): An Open-Source Platform for Automatic Sleep Staging of Rodent Electrocorticographic and Electromyographic Data.

Accurately capturing the temporal distribution of polysomnographic sleep stages is critical for the study of sleep function, regulation, and disorders in higher vertebrates. In laboratory rodents, scoring of electrocorticography (ECoG) and electromyography (EMG) recordings is usually performed manually by categorizing 5- to 10-sec epochs as 1 of 3 specific stages: wakefulness, rapid-eye-movement (REM) sleep, and non-REM (NREM) sleep. This process is laborious, time-consuming, and particularly impractical for large experimental cohorts with recordings lasting longer than 24 h, which are critical for the study of the circadian regulation of sleep. To circumvent this problem, we developed an open-source Python toolkit, Sleep Identification Enabled by Supervised Training Algorithms (SIESTA), that automates the detection of these 3 main behavioral stages in mice. We used a supervised machine learning algorithm that extracts features from the ECoG and EMG signals and autonomously scores recordings with a hierarchical classifier based on using logistic regression. We evaluated this approach on data collected from wild-type mice housed under both normal and different lighting conditions, as well as from mutant mouse lines with abnormal sleep phenotypes and from rats. We obtained mean F