Construction And Analysis Of Brain Functional Networks And Its Application In Depression

In the vast competitive pressures of modern society,the incidence of depression is very high.And currently,there is no medical examination or testing which can be used as a basis for the diagnosis of depression.The diagnosis of depression is mostly based on clinical diagnosis,and to a large extent this method of diagnosis depends on the physician’s clinical experience,and the accuracy of diagnostic is low.Therefore,how to improve the accuracy of diagnosis of mental illness is currently a problem of concern to doctors.Presently,a growing number of researchers are applying the complex network theory to the study of brain awareness.they regard a whole system of the brain as a complex network,and use more and more advanced technical means to collect data of human brain,and build brain functional networks using the data human brain they collected,use complex network theory to analyze and study the brain networks and achieve exciting results.Past research has proven:brain functional network of both patients with depression and the normal human had apparently small-world property and there were some differences between topologies of brain functional network of patients with depression and the normal human.Therefore,we can study the topologies of brain functional network of patients with depression and the normal human using complex networks theory.This article first build brain functional networks of all subjects in a continuous threshold value space using functional magnetic resonance imaging data of patients with depression and the normal human;analysis the network properties of all subjects’ brain functional networks using the knowledge of complex network basic heory such as the characteristic path length,clustering coefficient,degree,etc;And then,do statistical tests on global and local network properties of brain functional networks between normal group and depression group to find out the network properties which have significant differences between brain functional networks of the normal group and the depression group and to extract properties which could represent brain functional networks of patients with depression from different dimensions;Using different combinations of network properties which have most distinct differences as classification characteristics to classify all subjects using support vector machines algorithms,and set threshold to extract different number of global properties and nodal properties as classification characteristics depending on statistical significance to classify the depressive patients and normal peopleusing four kinds of classification algorithms,such as decision trees,Bayesian,artificial neural networks and so on.Through the classification study,the results found:If we use different combinations of network properties as classification characteristics,classification accuracy rate is highest using the combine of global and local properties as classification characteristics.If we use a different number of network attribute as.classification characteristics and different classification algorithms,the classification accuracy rate of support vector machines and neural network algorithm is respectively 82.78%and 81.36%,when the threshold p is 0.05.The studies of the diagnosis of depression in this article have some medical value.