Reliability of Family Dogs' Sleep Structure Scoring Based on Manual and Automated Sleep Stage Identification

Simple Summary: Sleep alterations are known to be severe accompanying symptoms of many human psychiatric conditions, and validated clinical protocols are in place for their diagnosis and treatment. However, sleep monitoring is not yet part of standard veterinary practice, and the possible importance of sleep-related physiological alterations for certain behavioral problems in pets is seriously understudied. Recently, a non-invasive electroencephalography (EEG) method was developed for pet dogs that is well-suited for untrained individuals. This so called polysomnography protocol could easily be implemented in veterinary diagnosis. However, in order to make the procedure more effective and standardized, methodological questions about the validity and reliability of data processing need to be answered. As a first step, the present study tests the effect of several factors on the manual scoring of the different sleep stages (a standard procedure adopted from human studies). Scoring the same recordings but varying the number of EEG channels visible to the scorer (emulating the difference between single channel versus four channel recordings) resulted in significant differences in hypnograms. This finding suggests that using more recording electrodes may provide a more complete picture of dog brain electrophysiological activity. Visual sleep staging by three different expert raters also did not provide a full agreement, but despite this, there were no significant differences between raters in important output values such as sleep structure and the spectral features of the EEG. This suggests that the non-invasive canine polysomnography method is ready to be implemented for veterinary use, but there is room for further refinement of the data processing. Here, we describe which parts of the sleep recording yield the lowest agreement and present the first form of an automated method that can reliably distinguish awake from sleep stages and could thus accelerate the time-consuming manual data processing. The translation of the findings into clinical practice will open the door to the more effective diagnosis and treatment of disorders with sleep-related implications. Non-invasive polysomnography recording on dogs has been claimed to produce data comparable to those for humans regarding sleep macrostructure, EEG spectra and sleep spindles. While functional parallels have been described relating to both affective (e.g., emotion processing) and cognitive (e.g., memory consolidation) domains, methodologically relevant questions about the reliability of sleep stage scoring still need to be addressed. In Study 1, we analyzed the effects of different coders and different numbers of visible EEG channels on the visual scoring of the same polysomnography recordings. The lowest agreement was found between independent coders with different scoring experience using full (3 h-long) recordings of the whole dataset, and the highest agreement within-coder, using only a fraction of the original dataset (randomly selected 100 epochs (i.e., 100 x 20 s long segments)). The identification of drowsiness was found to be the least reliable, while that of non-REM (rapid eye movement, NREM) was the most reliable. Disagreements resulted in no or only moderate differences in macrostructural and spectral variables. Study 2 targeted the task of automated sleep EEG time series classification. Supervised machine learning (ML) models were used to help the manual annotation process by reliably predicting if the dog was sleeping or awake. Logistic regression models (LogREG), gradient boosted trees (GBT) and convolutional neural networks (CNN) were set up and trained for sleep state prediction from already collected and manually annotated EEG data. The evaluation of the individual models suggests that their combination results in the best performance: similar to 0.9 AUC test scores.