The Study Of Resting-state FMRI Based On The Random Support Vector Machine Cluster

The relevant research results of brain science could effectively diagnose,prevent and treat neurological diseases.The government of each country began to attach great importance to the development of brain science and put it into the national strategy.Neuroimaging techniques such as the functional magnetic resonance imaging(fMRI)provide data support for the study of brain science.The researchers understand the brain's structure and operation mechanism through the functional connectivity of brain.In this paper,the brain functional connectivity of patients with Alzheimer's disease(AD)and autism spectrum disorder(ASD)is studied by using the random support vector machine(SVM)cluster based on fMRI.The following is the main content of this article:(1)A method of the random SVM cluster was proposed.Multiple SVMs were established by randomly selecting the samples and the features.The random SVM cluster combined by these SVMs was used for classification and feature selection.This method could achieve the purpose of dimension reduction to a certain extent and it is easy to implement.The prediction of this method was used to remove the redundant features,and the generalization ability was better.(2)A series of studies were conducted on AD patients using the random SVM cluster.The subjects including 25 AD individuals and 35 healthy controls(HC)were obtained from the ADNI database.The functional connectivity was made as feature of the random SVM cluster to classify the two groups,and the accuracy could reach to 94.44%.Furthermore,the 170 functional connections which are the biggest difference between AD and HC were found out.In addition,we could also find out some abnormal brain regions,such as the left orbital part of inferior frontal gyrus,the left precentral gyrus and the right anterior cingulate cortex.The study results show that abnormal brain regions in AD patients are universal.It is worth noting that the proposed random SVM cluster could be a new insight to help the diagnosis of AD.(3)A series of studies were conducted on autism patients using the random SVM cluster.The subjects including 45 ASD individuals and 39 typical controls(TC)were obtained from the ABIDE database.The graph metrics of functional connection network(degrees,shortest paths,local efficiency and clustering coefficients)were made as feature of the random SVM cluster to classify the two groups,and the accuracy could reach to 96.15%.Furthermore,the 272 features which are the biggest difference between ASD and TC were found out.In addition,we could also find out some abnormal brain regions,such as the inferior frontal gyrus,the hippocampus and the precuneus.The study results show that abnormal brain regions in autism patients are universal.It is indicated that the method of random SVM cluster may apply to the auxiliary diagnosis of ASD.