Resting State Brain Network Modeling Based On Functional Magnetic Resonance Imaging

Nowadays, the brain is the most complex, the most complete data processing system. It is a complicated brain network system which is formed by billions of neurons and the coupling between neurons. This paper uses resting state functional magnetic resonance imaging(functional magnetic resonance imaging, f MRI) and combined with complex network analysis method, in order to study on the brain functional network of generalized tonic clonic seizure patients(generalized tonic-clonic seizure, GTCS).Firstly, the resting-state f MRI data was preprocessed. The brain function network of epilepsy group and normal control group was constructed based on the method of Pearson correlation coefficient and Partial least squares algorithm. The attributes of networks were analyzed, including the average degree, shortest path length, clustering coefficient, global efficiency and the local efficiency. By calculating the network centrality index, we find that the praecuneus and cingulate gyrus are widely connected with many nodes, which are the key brain areas. The same method was adopted to model the brain function network of patients with epilepsy. Through inter-group analysis, the changed brain areas of epilepsy group compared with normal control group were found. The result showed that the "small world" property and network efficiency of people with epilepsy decreased, and the information transmission ability of the brain network was damaged; In addition, the brain areas which the correlation value increased significantly mainly concentrated in the thalamus and basal ganglia areas. It was associated with the generation and propagation of epileptic activity. The decreased brain areas mostly concentrated in the default network, especially the precuneus and prefrontal cortex, which suggested that the long-term repeated discharge of the whole brain may result in the default network damage.In this paper, the partial least squares algorithm and the traditional Pearson method are compared. The result shows that compared with the Pearson correlation analysis method, the PLS algorithm is better to construct the brain network model. It is not only expressed in the brain network threshold is generally high, the "small world" attribute is more obvious, but also the key brain regions that were inferred are more accurate and more consistent with physiological results. The results of this study shows that the Pearson algorithm and partial least squares algorithm can effectively infer the key brain areas of the normal control group and the brain areas of the lesions in patients with epilepsy, and it is of great significance.