Study On The Sleep Staging Of Rats Based On Eeg Brain Network

The sleep diseases have serious effects on the quality of sleep,and are harmful for human physical and psychological health.The diagnosis of sleep diseases is fundamental for sleep staging.Traditional sleep staging is mainly based on the subjective evaluation of sleep experts,which shows relatively low efficiency and sensitivity.Therefore,the development of an objective and efficient algorithm for sleep staging is urgent.In this context,automatic sleep staging algorithms based on electroencephalogram(EEG)signals are a main research focus in this field,which aim to extract direct features from EEG signals for effective sleep staging.Similarity to the brains of human and non-human primates,recent studies have reported that rodent's brain also has the default mode network(DMN).Importantly,it has been found that DMN has specific changes during different sleep stages,suggesting that the DMN characteristics might be a potential feature of sleep staging algorithms.In this thesis,we first analyzed the network variations of DMN and extracted the characteristics of DMN at different sleep stages by using the EEG signals recorded from the rats' DMN.Then,we combined these extracted DMN features with the support vector machine(SVM)to establish multi-classifiers for automatic sleep staging.Our main results are summarized as follows:1?We constructed the functional connectivity matrix of rats' DMN with the method of phase lock value to study the network variations in different sleep stages.Compared with the network topology structure of DMN during the sleep cycle,a significant change was observed in the theta band.Furthermore,we analyzed network properties of functional networks,and found that the network properties in the theta band remarkably changed during various sleep stages.2?Then,we employed three different classification features,including network properties,and the features extracted from original signals and weighted network based on the common spatial patterns,to identify the different sleep stages of rat.We found that the performance of the classification was worst for considering network properties as the classification feature.Furthermore,we compared the results derived from classification features extracted from the common spatial patterns,and found that the performance based on weighted network was better.Particularly,the performance was much better in the case that the number of the classification features was small.Together,our results indicate that the features extracted from weighted network based on the common spatial patterns can be used to automatically classify the sleep staging of rats.This finding not only contributes to relieving workload of human,but also provides a new perspective to promote the research of sleep staging.Besides,the present study might also helpful for the development of sleep monitoring device,clinical diagnosis and the treatment of sleep diseases.