Dynamic Functional Connectivity Pattern Analysis Based On Resting State Magnetic Resonance Imaging Of Brain

The human brain is the most complex structure in the world. People have been trying to discover the deeper mysteries of brain, and eager to unlock the mysterious veil. Functional magnetic resonance imaging(f MRI) provides a powerful tool for people to study human brain functions. Because of the large amount of information and simple operation, resting-state f MRI has been widely used. With the deepening research of brain functional networks, people have found the temporal variation characteristics of resting-state functional connectivity and paid increasing attention to it. Based on two groups of resting-state f MRI data, this dissertation explores the effect of schizophrenia and long-term driving behavior on the dynamic functional networks of human brains. The main contents and contribution of the dissertation are as follows:First, we built a methodology framework for the analysis of dynamic functional connectivity(d FC). The discovery of spontaneous neural fluctuations has made people begin to pay close attention to exchanges of the dynamic information between distributed brain regions. A growing body of studies have confirmed the temporal features of resting-state functional connectivity, and begin to apply dynamic functional connectivity to the analysis of brain functional networks. The second chapter has summarized the previous methods of feature extraction and pattern analysis in resting-state f MRI analyses, and develops a pattern analysis framework based on the dynamic functional connectivity as the basic features of neural activities. Then we apply this methodology framework to analyze resting-state f MRI data in the subsequent chapters.Second, we evaluated the possibility that abnormality of inter-network dynamic interaction could provide a potential biomarker for the diagnosis of schizophrenia. Based on dynamic functional networks, the third chapter attempts to identify schizophrenic patients from healthy controls with the methodology framework developed in the second chapter. The experimental results demonstrated that 81.3% of the subjects were correctly classified as either schizophrenic patients or healthy controls by leave-one-out cross-validation(LOOCV). Moreover, The statistical result of consistent features showed that the majority of the most discriminating functional connectivities were located across the intrinsic functional networks, indicating that the inter-network functional connectivities made more contribution for classification. Our results revealed significant abnormality in the dynamics of inter-network functional connectivity in schizophrenia, which contributed to the characterization and differentiation of schizophrenic patients, and may provide a potential biomarker for its clinical diagnosis. Finally, We estimated the potential influence on dynamic interaction between brain regions of long-term driving behavior. In the fourth chapter, the same method framework developed in the second chapter has also been used to study the effect of long-term driving behavior on the dynamic interaction between brain regions. The result of the classification showed that we could effectively differentiate drivers from non-drivers using the amplitude of low-frequency fluctuations(ALFF) of functional connectivity as basic features. 90.0% of the subjects were correctly classified as the drivers or the non-drivers by LOOCV. The majority of the most discriminating functional connectivities were located across the identified networks, indicating that long-term driving behavior had changed the dynamic information exchanges between specific brain regions. The brain regions within the vigilance network showed weaker dynamic interaction in the drivers relative to the non-drivers. Moreover, the drivers showed decreased dynamic interaction between the default and unimodal cortical networks, and between the fronto-parietal and unimodal cortical networks. Combining with the previous studies of resting-state functional connectivity, these results may contribute to explain the correlation between brain functions and driving behavior, and may have potential implications in understanding the plasticity of dynamic functional connectivity in resting-state brains.