Validation of Spindle Detection Algorithms to Find Association Between Sleep Spindles and Chronic Pain in Mice

Abstract

Sleep and pain are known to be intrinsically linked, and evidence suggests that non-rapid eye movement (NREM) sleep plays an important role in pain modulation. Sleep spindles, which are burst-like oscillations in the NREM sleep stage, are related to chronic pain regulation. Accurate sleep spindle detection is required to identify an association between chronic pain and the incidence of sleep spindles and contribute to a valid assessment of NREM sleep. In this study, we compare and identify spindles using four different sleep spindle detection algorithms, each known to find the spindles on par with the gold standard of spindle detection, manual spindle scoring, or visual identification. The primary aim of this work is to validate the automatic algorithms that are based on the sigma power parameters of humans for the detection of spindles in rodent EEG. Based on the F1 scores with the expert-scored spindles, we found that the automatic spindle detection by Algorithm 3 (Kaulen et al., 2022) had the greatest alignment with manual spindle detection in comparison to the others. The algorithms, however, were unable to distinguish between the Wild-type (WT) and Gi-DREADD KORcre type (Het) mice models with sufficient precision. This study highlights the need to validate spindle detection algorithms for rodent neural data to better understand the potential associations between sleep spindles and chronic pain.