Automatic Single Channel LFP Sleep Scoring In Mice

Sleep is a significant and basic biological function to all of creatures,the neuronal mechanism of sleep and the relationship between the sleep and the psychiatric disorders.Sleep scoring is the fundamental work for the research?diagnosis and treatments of sleep disorders.Clinical sleep scoring is generally done by experienced sleep researchers or experts combine with the software and their inspections,which is time-consuming and offline.There is urgent needed to design and develop a novel automatic sleep soring algorithm to assist the sleep study and the diagnosis of sleep disorders.Automatic sleep scoring not only reduce the burden of clinicians analyzing massive sleep data,but also facilitate low power and high-precision wearable sleep monitoring devices to come true.In this paper,we propose and implement a high precision sleep scoring scheme based on single channel local field potential(LFP).The main work of this article is shown as following:Firstly,we set up a single electrode LFP signal acquisition system to collect the LFP data.We used Cerebus neurophysiological signal acquisition system developed by Blackrock for data collection of LFP with multi-electrode chronic acquisition systems in mice during sleep.Secondly,we completed the offline TQWT wavelet decomposition of 16 levels for the single LFP signal,and 7 energy features based on wavelet decomposition sub-bands are extracted.By using the kernel density estimation and the Pearson correlation coefficient method,the extracted features are analyzed qualitatively and quantitatively,and the effectiveness of the features is also evaluated.Finally,we use a cascade of random forest ensemble learning classifier to achieve the automatic sleep scoring.The sleep auto scoring algorithm designed in this paper can achieve 90.8%,90.6%and 88%classification accuracy for three different sleep states of slow wave sleep(SWS),wakefulness(Wake)and rapid eye movement sleep(REM)respectively,and the overall accuracy can also reach 89.6%.The results of this paper show that the ensemble learning algorithm based on single channel LFP can be used in the realization of automatic sleep staging system.Provide the new method for the next online sleep staging for the brain-machine sleep-control system in the mice.