A Compound Classification Model for Schizophrenia Based on Brain fMRI and Network Modelling

Identifying of Schizophrenia patients' fMRI data has attracted lots of researchers' attention. Due to the complex nature of fMRI data, we need to generate “networks” to analyze the internal mechanism of the brain's function. It can be obtained by identifying the functional “nodes” and conducting a similarity analysis between the time series of the nodes. [1] gICA method can extract the “nodes” (ICA maps) of the network without prior knowledge of the brain's anatomy, and the Pearson correlation coefficient measures the similarity between “nodes”. With sliding window techniques, the dynamic model is built. In this paper, we propose a compound method which uses regression coefficient and correlation coefficient together as the input of the classifier to improve the experimental performance. The results confirmed that the proposed method is effective and generalizable.