Research On Spatial And Temporal Analysis Of Human Brain FMRI Dataset

The application of brain imaging techniques for research on the structural and functional characteristics of the brain plays a very important role in cognitive neuroscience. With the purpose of analyzing the spatiotemporal mechanism of brain function organization, the dissertation studies on functional magnetic resonance imaging (fMRI) datasets acquired in the unilateral finger-to-thumb opposition movement experiments, focuses on the data processing approaches, and investigates the spatiotemporal characteristics of neuron activities, of the activated regions and finally of the large scale cerebral functional network.Hemodynamic response and the related blood oxygenation level-dependence (BOLD) lay the physiological basis of fMRI. The dissertation presents a new data-driven sequential schema to study the block designed fMRI data. The scans in task blocks are investigated in a sequential manner. The nonlinear least-squares procedure is utilized to estimate the hemodynamic response function of each activated region. This schema expands the application of block designed fMRI datasets in the evaluation of hemodynamic response. By analyzing the movement experiment fMRI datasets, the spatial dispersion and the temporal difference between the hemodynamic responses of the activated areas including the contralateral sensorimotor area, the ipsilateral cerebellum and supplement motor area were testified. This study demonstrates the obvious spatiotemporal dispersion of the hemodynamic response.Grounded upon the functional segregation principle, a unified activation detection method, SPM-ICA is presented on the basis of Statistical Parametric Mapping (SPM) and independent component analysis (ICA). TICA (temporal ICA) is applied to fMRI datasets to disclose independent components, whose number is determined by the Bayesian information criterion (BIC). The resulting components are used to construct the design matrix of a general linear model (GLM) of SPM. Parameters are estimated and statistical inferences are made about regionally specific activations. SPM-ICA requires less priori knowledge than the conventional SPM, as well as circumvents the two shortcomings of ICA, namely lack of inference and lack of regional specificity. The adoption of BIC improves the determination of independent components from the posterior trial and error method to the priori unsupervised one. Results of Monto Carlo simulations and the analysis of receiver operating characteristic (ROC) curves demonstrate that SPM-ICA has higher statistical power and better performance.Moreover, the unified SPM-ICA method was applied to movement experiment fMRI datasets. Not only the traditional consistently task related activation areas, but also the transiently task related areas which corresponded to the posterior parietal cortex and frontal lobe were detected. The dissertation expands the definition of transiently task related components, and furthers the understanding of motor function and related brain areas.Grounded upon the functional integration principle, the dissertation presents the contextual functional connectivity analysis method (CFCA) and Bayesian functional connectivity analysis method (BFCA) on the basis of GLM functional connectivity analysis. CFCA combines GLM and the within-condition interregional covariance analysis (WICA), distinguishes states under different trials and tasks and can test functional connectivity strength under different states and its variance. This schema was applied to the movement datasets to evaluate connectivity strength and contextual connectivity relationship among region pairs of interest under four experimental conditions (left hand movement and baseline control, right hand movement and baseline control). The connectivity patterns of the associate parietal cortex and the ipsilateral BA (Brodmann Area) 6 and of the cerebellum and contralateral BA6 were recognized. BFCA utilizes Bayesian theory to estimate the functional connectivity strength between two regions and its significance. This method can explicitly evaluate the probabilistic certainty of the functional relationship, and by taking into account the with-subject variance to reflect the subject specificity BFCA can weaken the impact of outliers on parameter estimation, which makes it more robust.The dissertation studies the cerebral structural characteristics of the schizophrenics and investigates the motor function asymmetry. Voxel-based morphometry (VBM) was applied to schizophrenics'cerebral structure images. Significant decrease was identified in gray matter density of schizophrenics relative to the normal subjects, mainly in the right anterior cingulated gyrus, right superior temporal gyrus, and frontal lobe. Borrowing the idea of VBM, the dissertation presents a noval voxel-based multi-level functional asymmetry analysis schema, which explicitly defines the concepts of direct and indirect"inter-task functional asymmetry"and can automatically detect and evaluate different levels of functional asymmetry. This schema was applied to the movement fMRI datasets and some nuclei of basal ganglia (pallidum and claustrum) were found to have significant inter-task functional asymmetry, which demonstrated the importance of basal ganglia in motor function performance and manipulation.