Dynamic Brain Network Construction And Attribute Analysis For Working Memory EEG Data

Working memory is an important part of short-term memory.It is closely related to human cognitive activities.People's daily work,study and life are inseparable from its influence.Previous studies have found that many mental illnesses are directly related to working memory disorders.Studying the physiological mechanisms of working memory is of great help in exploring the generation and diagnosis of mental illness.In recent years,with the continuous development of related technologies,researchers have used various methods to study the physiological mechanisms of working memory,including Diffusion Tensor Imaging(DTI),Magnetic Resonance Imaging(MRI),EEG and MEG.Among these methods,EEG signals are characterized by high time resolution;and EEG data acquisition is simple and does not harm the human body.Therefore,EEG-based working memory research has gradually become a hot spot.There are many methods for dealing with EEG signals.Traditional analytical methods are mainly linear analysis methods,including time domain analysis,frequency domain analysis,and time-frequency analysis.However,EEG signals are nonlinear.Traditional linear methods do not take advantage of these features,so researchers have begun to focus on nonlinear methods.With the continuous development of relevant theories of complex networks,abstracting the brain into a complex network has gradually become a trend,and the EEG brain function network is increasingly accepted.The EEG brain function network usually uses electrodes as the transmitting nodes,and the correlation between the electrodes acts as edges.The traditional method of constructing EEG brain function network is for a continuous EEG signal,and the time overhead is relatively large,which is not conducive to largescale analysis.Therefore,this paper introduces the microstate concept into the brain network and constructs a microstate-based EEG brain function network.According to the research results,it can not only greatly reduce the time overhead of building a brain network,but also retain the main information in the signal,so that the network can more accurately reflect the state of the brain.In this paper,we constract a static brain function network based on microstates and a dynamic brain function network,and analyze the network topology and attributes.The data used in the study were 20 patients with schizophrenia and 20 normal subjects.The differential attributes of brain network were used as clinical diagnostic indicators.The main research contents of this paper are as follows:(1)A method for constructing a microstate EEG brain function network is proposed.The scalp electrode is used as a node,and the microstate time series is used instead of the EEG original time series to calculate the PLV as the edge of the network.Compared with EEG's traditional static brain function network in terms of time complexity,it is found that the time overhead is greatly reduced.(2)Construct a static brain function network based on microstates on EEG data oriented to working memory.Calculate the global attributes and local attributes of the network,and extract the stages,frequency bands,network attributes and brain locations that can distinguish between subjects with schizophrenia and normal controls.It is found that in the local attributes,the clustering coefficients and local efficiencies in the alpha and theta bands can be used as classification indicators.(3)Construct a microstate-based EEG dynamic function connection matrix on EEG data for working memory.The main connection matrix is obtained by PCA,and the related attributes of this matrix are extracted.According to the classification results,the classification accuracy of dynamic functional connection matrix is higher than that of the static brain function network.Explain that using dynamic methods can further differentiate patients with schizophrenia from normal people.