Research On Epileptic EEG Signals Based On Microstate

Epilepsy is a chronic neurological disorder caused by excessive and transient firing of neurons.Millions of people with epilepsy worldwide suffer from a reduced quality of life,loss of productivity,and possible premature death.In medicine,scalp or intracranial EEG reflects the electrical activity on the surface of the head and contains a large amount of pathological information,which is one of the important criteria for diagnosing epilepsy.At present,most methods such as time-frequency domain and nonlinear methods are used to study epilepsy EEG,and it is easy to ignore the spatial location information of EEG.The microstate analysis method can obtain the temporal dynamic information and spatial position information of the EEG,and is used in the research of many neurological diseases.However,it is unclear whether the results of EEG microstate analysis in patients are reliable,and there are very few studies on EEG and microstates in epilepsy.Therefore,from the perspective of microstates,this paper studies the reliability of epilepsy EEG microstate analysis results on the one hand,and analyzes the differences between epileptic and non-seizure EEG microstate parameters and the abnormal sequence attributes on the other hand.EEG signals of seizures,non-seizures,and healthy individuals were classified.The main contents of the research are as follows:(1)Reliability analysis of EEG microstates in epilepsyEEG microstate is a tool for analyzing multi-channel EEG signals.The number of EEG signal channels will affect the analysis results of the microstate.Although some studies believe that the number of electrodes will not affect the microstate,it only applies to the resting state of healthy people when their eyes are closed.The following four EEG microstates are studied.Whether the EEG microstates of epilepsy patients are also not affected by the number of electrodes remains to be studied.In addition,the analysis time of EEG signals also affects the results of microstate analysis.Therefore,this paper uses Intraclass Correlation Coefficient(ICC)and Coefficient of Variation(CV)to measure the reliability of EEG microstates with different electrode densities and different durations of data.The experimental results show that the EEG microstate analysis results are more reliable when the electrode density is greater than 18 channels and the data duration is longer than 36 s.(2)Epilepsy EEG microstate analysisUsing the microstate analysis method,the EEGs of 11 epileptic patients with and without seizures were studied.The peak value of Global Field Power(GFP)was extracted from the multi-channel epilepsy EEG signals,four classical microstate topographic maps were drawn,and the microstate parameters were calculated to observe the changes of parameter indicators,and statistical analysis was performed.Then,the microstate sequence is analyzed by information theory,including Shannon entropy,entropy rate,Hurst exponent,sample entropy,Lempel-Ziv Complexity(LZC),sequence stationarity test and Markov test.The experimental results show that the microstate parameters change when epilepsy patients have seizures,that is,the frequency of microstate B increases,the duration and coverage of microstate D increases,and the probability of transitioning from other states to microstate D increases(A-D,B-D,C-D).And there were significant differences in microstate parameters between seizures and non-seizures in epilepsy patients(P < 0.05).In addition,the information theory analysis results of the microstate sequence show that the Shannon entropy increases,the entropy rate decreases,the Hurst exponent increases,and the sample entropy and LZC decrease during epileptic seizures.With the increase of the microstate sequence,the sequence has stability at 30 s,and they all conform to the higher-order markov properties.It shows that the microstate sequence has more long-term memory,predictability and stability during epileptic seizures.(3)Classification of EEG microstate features in epilepsyIn order to use microstate features to classify epilepsy EEG signals,this paper not only calculated the traditional EEG microstate features of epileptic seizures and non-seizures,epileptic seizures and healthy people,epilepsy non-seizures and healthy people under different EEG rhythms,but also calculated the Hurst exponent features and time dynamic features(Autocorrelation Functions(ACF)and Aautoinformation Functions(AIF))of the microstate sequence are used to analyze the single features and several features are fused for classification.Experimental results: in the 1-40 Hz frequency band,the traditional features,Hurst index,and dynamic features of epileptic seizures and healthy people,epileptic seizures and non-seizures,epilepsy non-seizures and healthy people are significantly different,and the average accuracy of the fusion of the three features is significantly different.They are 95.0%,99.3%,and 99.3%,respectively,which are higher than the accuracy of single feature,two feature fusion and other frequency bands(Delta,Theta,Alpha,Beta and Gamma).This indicates that epileptic EEG microstate features can be accurately identified,and multi-parameter feature fusion in the 1-40 Hz frequency band can effectively improve the classification accuracy.