Analysis Of Children Brain Functional Connectivity Based On Electroencephalogram

The normal trajectory of childhood brain development is of great significance for the analysis of brain diseases.Many scholars study the brain functional connectivity of normal children at different ages based on EEG,but usually adopt traditional coupling methods and connectivity analysis methods,which will affect the reality and comprehensiveness of the representation of brain connectivity.On the other hand,people often ignore the analysis of nonassociative features of electroencephalogram(EEG)during the analysis of functional connectivity,which may lead to missing potentially important information.Based on the non-rapid eye movement(NREM)EEG of 0～17 years old children,this paper studies the differences in brain functional connectivity of children of different ages.The innovation points are mainly as follows:(1)Network binarization and deep learning methods were used to solve the influence of EEG noise and individual differences on the reality of brain connectivity representation.(2)Apply complex network analysis method and propose new network measurement.The main work and achievements are as follows:1.For dealing with the problem of low signal-to-noise ratio of EEG and the individual differences,this paper proposes entropy stability based threshold searching method to construct individual level functional connetivity(ILFC).This method is used to eliminate personalized connections and noise-generated connections in functional networks.Through the analysis of functional separation and functional integration of the network,it is found that:(1)The β band of EEG is the key frequency band that reflects the change of functional network separation and integration characteristics with age.(2)The brain changes from network separation to network integration during 3 months to 3 years of age.Then,in order to study the network centrality of age groups,group level functional connectivity(GLFC)was established based on voting method.Then variance of overlaying node degree(VD)is proposed on the basis of node degree,which reflects the prominent degree of brain network center.The results showed that this index was negatively correlated with the aging,which could reflect the maturity of the brain.2.A weighted functional connection construction method based on one dimension convolutional network(1D-CNN)and canonical correlation analysis(CCA)is proposed.This method reduce EEG noise and individual information before the connection construction stage.Non-parametric test was used to analyze the main functional regions where the connection values changed at different age stages.Then,the 1D-CNN model was used to study the differences of EEG uncorrelated features at different ages,and the main functional regions where EEG uncorrelated features change at different ages were obtained.Based on the above experimental results,it is found that:(1)The functional connections of the brain that change with age are mainly cross-regions functional connections.(2)For each brain functional region,the extent of functional connectivity and EEG non-correlation characteristics change with age is usually inconsistent.(3)3-year-old is the critical stage of children’s cognitive development.3.The children’s brain functional connection analysis system is developed.Based on Matlab GUI software,the functional connection construction method and network analysis method mentioned in Chapter 3 are applied to this system,which can realize three functions:EEG data preprocessing,functional connection construction,and comparison analysis of network connection and topological features.The system can assist doctors to study and analyze the brain functional connectivity of individual children.