Research On Functional Brain Networks And Its Dynamic

Until now, the human brain is known as one of the most complex and most delicate system, which is organized into parallel, interacting systems of anatomically connected areas. The movement, memory, language, thinking, emotion and other cognitive activities are controlled by the human brain. With the development of imaging techniques, especially the mature of functional magnetic resonance image technology, the brain and its activities can be continuous observation overall situation, complex network technology which based on graph theory has become a powerful tool for neuroscience research. Combined with imaging techniques and complex network theory, more and more studies explored the brain mechanisms from the network perspective. From the system level, brain network can study the mechanism of brain connections, and then to reveal the internal organization model of brain. Many studies have revealed the mechanisms and inherent principles of brain disease from different levels used by the brain network methods. The current work dedicated to investigating the modeling and analysis methods, theory, application and dynamic process of brain network, by means of functional magnetic resonance imaging (fMRI). Three aspects of this dissertation have been put forward:The first part is investigation of related issues in the functional brain network construction, including data noise reduction and different connection methods. In this part, we highlighted a data noise reduction method of resting-state fMRI datasets, and explored the different connection methods of functional brain network construction.First, noise and individual differences arise from disturbances in the effective use of resting-state fMRI datasets. For the first time, this dissertation has proposed the point process to reduce the noise of resting-state fMRI datasets. The datasets of healthy controls and patients with diabetes were used to verify the methods. The results illustrated that differences between the healthy controls and the patients were more obvious in signals treated with the preprocessing point process. SVM classification results showed that higher recognition accuracy was achieved after preprocessing of the point process. These findings may suggest that the point process approach can reduce BOLD signals noise, provide a new method for functional magnetic resonance data preprocessing, and may act as a method for early data preprocessing of computer-aided disease diagnostics.Additionally, there are different connection methods in studies of brain network, however, the differences between the different connection methods were very few systematic studies. Using resting-state fMRI datasets of healthy controls, this dissertation compared the different connection methods. Based on brain activity of motion control, the time courses of related brain areas were extracted in this dissertation. Then, different methods were used to build the connection between the seven brain areas. The results illustrated that different connection methods can affect the results in the research of brain network. In actual research, connection methods should be chosen with the actual experimental conditions and purpose. Meanwhile, wavelet transform coherence (WTC) method was firstly used to construct the functional brain network, and, the reasonableness of the WTC method for functional brain network construction was verified by the actual data. The results showed that the brain network in different bands exist large differences, this mean that it is essential to observed the functional brain network from different filter bands.In the second part of this dissertation, we explored the construction methods of local and global brain network. This section focused on the connection changes of main motor areas in stroke patients, as well as global functional brain network changes in the Alzheimer’s patients in different frequency bands.First, stroke is a common diseases in recently years, and motor dysfunction was the most prominent symptom for stroke. However, very few studies explored the connections change of movement areas after stroke. This dissertation was firstly to investigate the difference of functional network connectivity (FNC) between stroke patients and healthy controls with fMRI, and to explore the possible alterations in brain network after stroke. Six spatially independent components highly correlated to the paradigm of experiments were extracted using ICA (independent component analysis) as the nodes of network, then, the functional network connectivity was used to explore the weaker temporal relationship among them. The results of network modeling demonstrated that the network connectivity of patients was much more complex than that of controls. More importantly, we found some compensation circuits produced after stroke. These results implied that the method of FNC and ICA were effective way for neuropathy research, and may provide indicators for the assessment of patients with stroke rehabilitation process.Secondly, Alzheimer’s disease (AD) is a common disorders in the elderly, previous studies have indicated that the cognitive deficits in patients with Alzheimer’s disease (AD) may be due to topological deteriorations of the brain network. However, the results of studies on brain network of AD existed inconsistencies. This dissertation investigated large-scale topological properties of moderate AD patients and healthy controls at three distinct frequency bands (0.01-0.06Hz,0.06-0.11Hz, and0.11-0.25Hz). The results showed topological properties of brain networks of AD change with different frequency bands, such results may explain why the previous study on functional brian network of AD were inconsistency. We found the global efficiency, small-world property, the clustering coefficients of AD patients were decline, the assortativity coefficient was bigger and the synchronous was weakened in low frequency bands. The results may be the biological identity of the AD early diagnosis. Also, the results provided fundamental support for optimal frequency selection in future related research.In the third part of this dissertation, we explored and analyses brain network evolution model and dynamic process. We firstly proposed the evolution model of functional brain network in the disease and the aging process, developed the brain network analysis method which is based on information theory.First, for it is very difficult to effectively observe the slow changes of nervous system used by the currently technology, we cannot obtain exhaustive datasets in the studies of nervous system. This dissertation set up two different brain network evolution models to simulate dynamics process of brain networks under different circumstances by using the resting state fMRI data. In AD evolution model, we firstly regarded the network betweenness and anatomical distance as control factors. Meanwhile, for the first time, we introduced the differential evolution algorithm to optimize the brain network control factors in the process of evolution in age evolution model. We have got good results in both two evolution models, the reliability of the models were verified by support vector machine and separate sample datasets respectively. As far as we know, it’s the first time to study the function changes of the human brain by using the computer simulations at the network level. Our study may provide more convenient research methods for the study of senile diseases such as Alzheimer’s disease.In addition, the human brain is the organ of information processing, however, current studies on brain network are rarely based on the perspective of informatics, and there are few corresponding network indicators based on informatics to evaluate the brain networks. This dissertation calculated the random entropy and fractal dimension of time series of resting-state fMRI datasets of the young and the old based on symbolic dynamics, also calculated the whole brain information entropy. The results illustrated that symbolic dynamics can reduce the noise of fMRI datasets. We also found the probability entropy of frontal regions, basal ganglia area increases in the old group, this indicated that it is more difficult for these two regions of older group to search for information, and, this may be an important cause for mental retardation in the elderly. The results also illustrated the information entropy theory is an effective indicator to evaluate brain function network. This study provides new ideas for explain the brain mechanisms, and provides a new way for fully understand the brain mechanism.