Study On Eeg Micro-state Analysis And Recognition In Sleep Deprived Patients

Sleep is essential for the brain to recover and keep it functioning optimally.Sleep deprivation(SD)interferes with the normal functioning of the human brain and even impairs brain function,usually resulting in reduced response times,reduced alertness,decrease the perceptual and cognitive ability.The cognitive effects of total sleep deprivation(TSD)on the brain remains poorly understood.Electroencephalography(EEG)is a very useful tool for detecting spontaneous brain activity in the resting state.Quasi-stable electrical distributions,known as micro-states,carry useful information about the dynamics of large-scale brain networks.This article mainly used micro-state analysis method,to analyze the resting state EEG data of 24 healthy volunteers in rested wakefulness(RW)and TSD,explore SD-related bio-markers,and study abnormal brain activities were caused by the sleep deprivation.A support vector machine aided classification model was established based on the Micro-state characteristics data.The SVM parameters were optimized by MVO algorithm,and a high accuracy was obtained for the diagnosis of the effect of sleep deprivation on brain activity.Therefore,a new idea has been proposed to diagnose whether sleep deprivation affected human brain activity.The main contents are as follows:(1)EEG signals were preprocessed before and after 24 hours of sleep deprivation.EEG signals are bio-electrical signals,which will undoubtedly be mixed with various noises.In this thesis,independent component analysis(ICA)method was used to eliminate the noise of ECG,EOG and EMG mixed in EEG signals,so as to obtain relatively clean EEG signals for subsequent research and analysis.(2)EEG microstate analysis was performed before and after sleep deprivation for 24 hours.The EEG of 24 subjects before and after sleep deprivation was studied by micro-state analysis.Six typical micro-state templates were extracted from EEG signals.The time characteristics of different micro-states and the state transition probabilities between them were analyzed statistically.The clustering results of micro-state were consistent with the previous studies.Statistical analysis of the time characteristics of six micro-states showed that the global explained variance(GEV),occurrence frequency and time coverage of micro-state A decreased significantly after sleep deprivation,while the global explained variance,occurrence frequency and time coverage of micro-state D increased significantly.Moreover,subjective sleepiness was significantly negatively correlated with micro-state parameters of class A,and positively correlated with micro-state parameters of class D.Transition analysis revealed that class B exhibited a higher probability of transition to class D and F in the TSD compared to RW.These differences can provide effective evidence for computer-aided diagnosis of the effect of sleep deprivation on brain activity.(3)A feature classification method of EEG micro-state before and after sleep deprivation was proposed,which was based on support vector machine with parameter optimization.The selection of error penalty parameter C,kernel function and its parameters affected the classification performance of SVM.In this paper,a parameter optimization scheme was proposed,and the Multi-Verse Optimizer algorithm and cross validation method were used to optimize the parameters.This method effectively avoids the local optimization problem of traditional optimization methods.And the experimental results showed that MVO-SVM method had a good classification effect(95.26%).