Research On Classification Of ADHD Based On The Characteristics In Brain Activity-network

ADHD is a very common psychotic disorder in children. Its typical symptoms include inattention, hyperactivity and impulsive behavior. ADHD research has been a serious public health problem in our country and even in the world for more than ten years. Because so far the etiology and pathogenesis of ADHD is not clear, the diagnosis and treatment is difficult. The objective diagnosis and effective treatment of ADHD has become one of the hot research subjects now. In order to study the objective diagnosis of ADHD, we have done the following research:First, we introduced several kinds of methods which can be used to analyze brain activity network based on the resting state fMRI (function connectivity analysis, ReHo analysis and ALFF analysis).Second, in order to classify109subjects included37ADHD children and72normal developing children(data were selected from a large sample RS-fMRI dataset publicly released as "The ADHD-200Sample" in the "1000Functional Connectomes Project"), we extracted three features, such as functional connectivity, ReHo graph and ALFF graph, then we reduced the number of features to the200functional connections (generated from t-test) most reliably different between ADHD and normal children in each round of leave-one-out cross-validation. Then we introduced these200features to SVM classifier to explore the effective classification of ADHD and normal children with leave-one-out cross-validation. However, the classification results are not satisfied. The best classification rate is only66.06%.Third, we attribute the unsatisfied results to something wrong with feature extraction before the classification. In the light of unsatisfied classification results based on the features of functional connectivity, ReHo graph and ALFF graph, we try to apply features of brain functional network independent components to classify subjects. As one part of this target, the important role of intelligence quotient (IQ) for resting state brain functional network, especially parieto-frontal network (PFN), was researched. In view of some previous studies suggested strong relationship between intelligence and brain network PFN (parieto-frontal network) in adults, we study whether this relationship exist in children. We performed independent component analysis of resting state fMRI data of84children and50adolescents separately, and then correlated full-scale IQ with the spatial maps of the bilateral PFNs of each group. And the results showed that the relationship exist in the participants. This suggests that extracting features of brain functional network (such as PFN) independent components in the classification must take into account the influence of the intelligence quotient. Eliminating the influence of IQ on the classification is one of the important contents for further research on ADHD classification in the future.