Microstates Study On Resting-state EEG In Abstinent Heroin-Addicted Individuals

Heroin addiction is an important healthy problem worldwide,and its difficulty in withdrawal and high relapse rates have plagued addicts and researchers for many years.The current study was dedicated to exploring more objective physiological indexes for the diagnosis and treatment of heroin addiction.On the one hand,with the understanding in the mechanism of heroin addiction,we hope that we could explore the relationship between heroin addiction and brain damage,as well as the degree of addiction and the severity in brain injury from the perspective of neuroscience,which is primary for proposing more objective parameters.On the other hand,if we could propose a more efficient and objective parameter system,then the parameters would be the key indicators for clinical application of diagnosis and treatment for heroin addiction.This research focused on exploring the difference in microstates of eye-closed resting states electroencephalogram(EEG)signals between abstinent heroin-addicted individuals(AHAIs)and healthy controls(HCs),as well as the effectiveness of distinguishing the two groups using the microstates parameters.This was a research on topological analysis based on EEG microstates and binary classifier design.Compared to the common source localization analysis methods,the analysis of microstate could indicate the time domain information of EEG signal.Compared to traditional EEG signal analysis,this method preserves as much spatial resolution as possible.That is,the method has both high time-frequency resolution and spatial resolution.In addition,three new types of microstate features were proposed in this paper,which provided the possibility of connectivity research for microstate analysis.After calculating the 128 microstate characteristics of the closed-eye resting EEG signals of 50 subjects in the abstinent heroin addicted individual and the healthy control group,The binary classification was found to be the optimal classifier and the average classification accuracy reached 72.3%,which also confirmed the validity of the microstate parameters as a physiological indicator for the diagnosis of heroin addiction.It is worth noting that in this study,we also tried to weight the featuresusing the GA algorithm and then used the weighted features for classification.By comparing the results of the four feature weighting methods of GA,CE,CV and Critics,it was found that the accuracy was effectively increased to 81% by the GA weighting method.The current study was consisted of two parts.The first part was based on microstate analysis,combined with statistical methods such as repeated measures analysis of variance,T-test and randomization test(permutation test),to explore and test whether microstates differ between AHAIs and HCs more scientifically,and figured out whether the transitions from a preceding state into a next state occurred randomly.The second part explored how to construct actual clinical applications,namely the design of binary classifiers based on microstate parameters.For the feature extraction,three more types of parameters were found in our research than the parameters commonly used in the analysis of microstate.For the design of classifier section,this paper used genetic algorithm(GA)to weight the optimal features.The feature was selected by sequential floating forward selection(SFFS),and the feature subset was evaluated by the wrapper type evaluation function.At the same time,the optimal classifier was also elected by wrapper type evaluation function.The results of this study was benefit to understand the mechanism of heroin addiction,and to confirm that the microstate parameters could provide objective and efficient evaluation criteria for the diagnosis of heroin addiction.We believe that our study will be of great benefit to the development of feature diagnosis and treatment for heroin-addicted individuals.