Research On Classifications Using Abnormal Brain Network Topological Metrics In Major Depressive Disorder

Under the influence of physiology, psychological as well as social environmental factors, the human brain show function disorder which lead different degree of hurdle disease to the acts of cognition, emotion, willingness and other mental activities. We call it depression.Currently, the diagnosis of anxiety and depression is still based on external symptoms demonstrated by the patient to diagnose whether the patient is sick. Brain network is a tool widely used for identifying topological properties of the exception, and for the diagnosis of nervous system diseases with a new perspective. However, it remains unclear whether resting state functional brain network topological metrics can be used in machine learning methods for the classification of MDD. Further research is required to clarify the most appropriate feature selection methods, to establish the classification model, provide support for computer aided diagnosis of patients with depression. The main work as follows: During the study, we collected resting state magnetic resonance imaging data from38drug-naive, first-episode MDD patients and28normal controls to construct resting state functional brain networks.In addition, we calculated the Aggregation coefficient, the shortest path length, the degree, betweenness centrality and nodal efficiency using graph-theoretical methods, the typical brain network topological properties as the classification features, using feature selection method for the selection of the topological properties.Classifier is constructed by five different classification algorithms, classification including support vector machines, neural networks and decision trees. The statistical significance of features was used as the threshold for filtering features so that the performance of classifiers with different numbers of features could be evaluated.With the research to the28characteristics (P<0.05), support vector machine (SVM) based on radial basis function neural network algorithms and algorithm receiving the highest average accuracy (84.36%and80.70%, respectively). Studies have shown that depression and abnormal brain function related to network topology metrics, and the statistical significance of a node metric can be successfully used in feature selection of classify algorithm. The construction of the classify model of has a certain reference value for diagnosis of depression.