A Study on the Classification of Schizophrenia Based on Single Gray Matter Networks

In order to study the covariance of subtle differences in gray matter in schizophrenic patients, single subject brain network model was constructed by using voxel similarity index. The gray matter was divided into fixed size cubes, the single subject gray matter covariant network was constructed by means of Pearson correlation between voxels, and the network was mode sparse by empirical null model method and finally normalized to a 90×90 gray matter covariant network. The results show that the topological properties of the single subject gray matter network of schizophrenia patients were significantly changed, the node efficiency and degree of the parietal lobe and frontal lobe were different, and the efficiency of the right precuneus was significantly correlated with the scale. Support vector machine, KNN, Naive Bayes and random forest algorithm were used to establish the classification model of schizophrenia patients, whose classification accuracy was significantly higher than the classification results based on the brain region. In conclusion, the personal gray matter network based on the gray matter similarity of voxels can better detect the abnormal brain network in schizophrenia, and can be applied to the early diagnosis of schizophrenia.