Research On Dynamic Functional Connectivity And Classification Based On Default Mode Network Of Epilepsy Patients

Epilepsy is a chronic brain disease characterized by recurrent seizures,which seriously affects human physical and mental health.Exploring the association between brain functional connectivity and the disease has always been a hot field.Using functional connectivity to study the disease can analyze the functional changes of patients,help to understand the pathogenesis of mental illness,and propose effective treatment measures.Traditional functional connectivity can’t reveal the temporal changes and dynamic interaction patterns of the brain.At the same time,disease classification based on functional connectivity also faces the problem of low classification accuracy.The current research is devoted to the dynamics of the brain of epileptic patients and the improvement of classification accuracy.Therefore,whether the study of dynamic functional connectivity(d FC)of epilepsy can provide help for the understanding of the disease deserves further study.In order to solve the above problems,this paper uses resting-state functional magnetic resonance imaging(fMRI)as a data set to construct dynamic functional connectivity to explore the characteristics of Default Mode Network(DMN)of Juvenile Myoclonic Epilepsy(JME)patients,and realizes the classification of JME patients through dynamic functional connectivity feature extraction.The main contents of this paper are as follows :(1)Information acquisition of DMN in JME patients.Firstly,the data is preprocessed according to the characteristics of the currently known fMRI data,and then the data is denoised by the group independent component analysis algorithm for the data mixed with the noise source,and the components whose peak coordinates are mainly located in the gray matter and have a low level of correspondence with blood vessels,ventricles and artifacts are retained.Using the existing template,the brain regions in 8 DMN such as the left anterior cingulate gyrus were selected as the regions of interest of the study,which provided the basis for subsequent research and analysis.(2)DFC analysis of default mode network brain regions in JME patients.The d FC between 8 brain regions obtained by independent component analysis was constructed by sliding window method.Four different d FC states were obtained by K-means clustering,and the d FC indexes in each state were calculated.Finally,the difference indexes between JME patients and normal control group were extracted,and the correlation analysis was performed with the score of National Hospital Seizure severity scale.The results show that the d FC study of DMN in JME patients can provide important clues for revealing the pathogenesis of JME.(3)Study on the classification of JME patients with default network dynamic functional connectivity.Firstly,four different brain functional connections are constructed,and the four brain functional connections are used as the eigenvalues of classification.Then,the dimensionality reduction of eigenvalues is completed by principal component analysis,and the parameters of support vector machine are optimized by whale optimization algorithm.Finally,the four different eigenvalues after dimensionality reduction are classified.By comparing the accuracy of the classification results,the best feature extraction method for the classification of JME patients is obtained,which effectively solves the problem of low classification accuracy of JM E patients using fMRI.