Assessing accelerometry as an alternative to electromyography for mouse sleep scoring

Abstract

For sleep scoring in rodents, the use of electromyography (EMG) is considered the gold standard to assess muscle tone; however, EMG electrodes require invasive surgical implantation, are sensitive to muscular placement, can have extensive noise due to non-muscular signal contamination, and are at risk of failure due to mechanical stress. The use of an accelerometer (Acc) integrated on the recording headstage offers an alternative for movement detection. However, the degree of agreement between these two approaches has never been systematically evaluated. To understand if these tools allow similar scoring outcomes, seven trained scorers manually classified mouse sleep recordings using hippocampal local field potential (LFP) together with EMG or Acc. When comparing scoring between EMG and Acc, overall epoch-by-epoch accuracy exceeded 90% with a kappa score above 0.8, indicating excellent agreement. However, Acc was found to under classify Wake by 4.92% while overclassifying NREM sleep by 3.29%, possibly reflecting a decreased sensitivity of Acc to subtle changes in muscle tone used in distinguishing Wake from NREM sleep. Comparing EMG and Acc signal for agreed epochs vs disagreed epochs revealed a weaker correlation between the two signals for disagreed epochs, especially for Wake and NREM, suggesting that disagreements reflected genuine signal differences rather than scorer subjectivity. These findings support LFP/Acc as a valid approach to sleep scoring, however, it is noted that Acc based scoring slightly underestimated Wake epochs, favoring NREM sleep, relative to traditional EMG approach.