Brain Functional Network Research Of EMCI And AD Based On Resting State FMRI

The brain is a complex system of interactions between different brain regions activities.The study of brain functional networks, on the one hand, help to explore the mechanism of the brain neural activity,on the other hand,it has important application value and significance to finding the pathogenesis of brain related diseases. Alzheimer’s disease is a neurological disease. At present, there is a lack of effective measures and treatment. Early mild cognitive impairment is a high risk of Alzheimer’s disease.Reseach on their brain functional networks provides an important basis for early warning and the clinical diagnosis of Alzheimer’s disease.This paper analyses 38 normal controls(NC), 25 Alzheimer’s disease(AD) patients and 42 participants with early mild cognitive impairment(EMCI) resting state functional magnetic resonance imaging data, and with the complex network theory to analyze the differences between their brain functional networks. Finally, according to the network parameters, this study made a preliminary classification and prediction for three groups of subjects by support vector machines. The specific studies are as follows:(1) Brian functional network modeling and threshold selection strategy. Pearson partial correlation is used to measure the functional connectivity between brain regions, and the positive and negative correlations are taken into account. This paper uses relative threshold method and absolute threshold method, and the effect of the two methods are compared.(2) Network index calculation and analysis. The study calculates the two global network properties and four network nodes properties. Based on global network properties, we analyze and compare the small-world network nature of three group subjects. The paper calculates the area under the curve of node attributes at equally spaced thresholds, and the integration result is used as the classification features.(3) Classification prediction and feature extraction. In order to obtain the network parameters with significant differences in brain areas, this study uses t-test and k-s test method for feature selection. Diseases has been identified under SVM. The paper uses grid search method for parameter optimization, and then computes leave one cross validation classification accuracy.The results show that, no matter which threshold strategy is selected, compared with the NC group,the small-world properties of AD patients has declined, while the small-world nature of EMCI group is enhanced. Under ideal circumstances,the classification accuracy are above 80%. From the classification effect of view, relative threshold method is better than the absolute threshold method, and the classification accuracy based on t-test is generally higher than k-s test results...