The Classification Comparative Study Of Depressive Patients’ Feature Of Brain Network Properties

Functional magnetic resonance imaging technology is the human brain blood oxygen level of reliance on the brain research. The technology with its high resolution on the body’s non-invasive, image and is widely used in various studies on the human brain. Currently at home and abroad with functional magnetic resonance imaging technology research focuses on the cognitive domain of the human brain, including the understanding of things, memory, and to identify such. The research is mainly carried out through the statistical analysis of functional magnetic resonance imaging data.With the continuous development of the field of artificial intelligence, a growing number of machine learning algorithms have been proposed and applied to the actual study, but in the existing research in machine learning algorithms applied to human brain functional magnetic resonance imagingstudy. Construct functional brain networks using functional magnetic resonance imaging of patients with depression and normal controls, the specific characteristics of the network properties of brain areas and functional brain networks for classification and comparison of discriminant classification algorithm for patients with depression and normal. The main work is as follows:Functional brain networks constructed using the depressed patients and normal functional magnetic resonance imaging data to construct the corresponding functional brain networks; to be studied the selection of brain regions, using non-parametric permutation test to analyze the network properties of the90brain regions obtainedsignificant difference is relatively large areas of the brain as the study of brain regions; feature selection, to choose to treat the study network properties of brain regions to get the characteristics of four typical network properties, namely the middle of centrality, the aggregate coefficient, the shortest path length;classification algorithm in the classification of fMRI data, highlighting the Naive Bayes and support vector machine classification algorithm, SVM (support vector machine) classification algorithm to treat brain regions of four typical network properties characteristic classification and compared them for each group.Selected patients with depression and normal brain range of functional brain network classification of the property, the classification accuracy rate of the ten selected brain regions of depression and control groups in more than72%, among which the left supplementary motorDistrict and is responsible for learning and memory, the right side of the hippocampus in classification accuracy, reaching more than80%. In order to further reduce the impact of data between different qualitative data analysis results, the data were standardized, and standardization of data re-analyzed the results of the comparison is not significantly improved, its value is still essentially between70-80%, the highest classification accuracy of brain regions is the right lentiform putamen, the accuracy rate of82.6087%. Both the classification accuracy are much higher than the random classification accuracy.This clinical diagnosis with a certain reference value.