Functional segregation and functional integration are two basic organizational principles of brain functions. Based on the functional connectivity resting-state functional magnetic resonance imaging (fMRI), we explored these principles in two different topics. The first topic of the dissertation is about function integration, investigating the properties of brain networks. The second topic is about functional segregation, developing approaches for the brain parcellation. The main contents of the dissertation are as follows: Ø We investigated the functional architecture of the brain network of monkeys and compared it with those of human adults and children to illustrate patterns of brain evolution and brain development. Functional connectivity density (FCD) mapping based on the resting-state fMRI has recently been used to locate cortical hubs of the adult brain. In previous studies, only one threshold was defined for the significant functional connectivity, since FCD maps of adults are relatively robust under different thresholds. In the present study, we examined the functional architecture of the monkey, human child (7-11 years old) and human adult brains, and compared our results with former related studies on infants. We present results showing that FCD maps of children change qualitatively with different thresholds reflecting the level of cerebral maturity. Compared with the adult brain, the lateral prefrontal cortex and areas related to the default mode network of the child brain show more differences in the FCD than the primary sensory and motor areas and insular cortex. When compared with results of monkey, we discovered the difference between monkey and human in FCD maps pattern. In addition, we investigated the connectivity-based segmentation of the cingulate cortex and brain networks of monkeys. By comparing these results with former related studies in humans, we have a better understanding of the patterns of human brain development as well as brain evolution. Ø We proposed approaches for feature reduction and selection for the functional-connectivity-based brain parcellation and developed semi-simulated data to evaluate these feature reduction approaches. Traditionally, a fundamental field for neuroscience has been the segmentation of structurally and functionally distinct subunits of the brain utilizing invasive techniques and post-mortem investigations based on cytoarchitecture and anatomical landmarks. More recently, resting-state functional magnetic resonance imaging (rs-fMRI) has been used to perform the parcellation of brain regions based on the information available in functional connectivity maps. However, brain voxels are not independent units and adjacent voxels are always highly correlated, leaving functional connectivity maps containing redundant information. This redundancy not only impairs the computational efficiency during clustering, but also reduces the accuracy of these results. In the present study, we proposed and investigated two feature reduction approaches that can reduce this redundancy: (1) based on Principal Component Analysis (PCA) and (2) on Affinity Propagation Algorithm (AP). These two approaches were tested for their feature reduction ability during the parcellation of three brain regions of different sizes. In addition, we developed a method of constructing semi-simulated data from real fMRI data to evaluate these two approaches. With the known ground truth of the semi-simulated data, conclusions drawn from comparison between different approaches would be more reliable. These results suggested that a feature reduction on functional connectivity maps is both feasible and necessary.
修改评论