Based On Magnetic Resonance Brain Network Analysis Method And Its Applications

Being the command center of a human body, brain has a complex organization in both structure and functional mechanism. Generally, functions among several brain regions may be similar, which means these regions form a network, usually called static network. On the other hand, interaction occurs between some brain regions when task performance, such as modulation and inhibition, forming a dynamic network. With the progress of the times, the human being not only explores his living environment, but also makes a good study of his own body, especially the brain. In this case, brain mapping, especially the invention of magnetic resonance imaging (MRI) technology, is a valuable tool for human to study their brains. Recently, the brain network, based on the research of the MRI,has been widely studied. The static and dynamic networks, we mentioned above, are the most two important and well known networks in this field. These researches not only are helpful to study ourselves, more importantly, they play a important role in disease study. In this paper, based on the magnetic resonance imaging data set, we consider the methods in brain network analysis and their applications. The numerical results demonstrate the effectiveness of our methods.Three aspects of this dissertation are as follows:1. Propose a Bayesian study based radial basis function(RBF) neural network for locating the function networks when task performances. Several brain networks, such as visual, imaginary and hand move, are well separated, which indicates the effectiveness of this method in the localization of function networks. It also demonstrates that the method can discriminate the responses of different tasks. That means, this research provides us a new tool for function localization when multiple tasks performances.2. Combining the functional and the structural connectivity to study the default mode network of patients with mesial temporal lobe epilepsy, we find the functional connectivity and the structural connectivity between PCC and bilateral hippocampus are both decreased. Moreover, both the functional connectivity and the structural connectivity are significantly positively correlated, which means that combining the functional connectivity and the structural connectivity is effective and reliable in MRI data processing, and thus provides us a new way in brain research .3. First of all, we have an introduction of the Granger causality which is usually used to investigate the dynamic network. Meanwhile, a modification of causality analysis method, called kernel causality, is proposed since the granger causality fails to handle the cases when detecting nonlinear causality correlations among MRI signals. The simulations show that, compared with granger causality, kernel causality can easily detect the nonlinear causality between simulate signals. Moreover, making use of kernel causality on MRI data processing, we can obtain a quite better characterization of the information communication in human brain.