Research On Functional Parcellation Of The Human Brain

The information exchange in human brain at multiple spatiotemporal scales is the basis for supporting its complex cognitive and behavioral functions.We try to study the neural information interaction mechanism of human brain by magnetic resonance imaging,a non-invasive technique with high temporal and spatial resolution.In this dissertation we mainly focused on the functional connection of the human brain at resting state,and studied the functional unit division and interaction mode at multiple spatiotemporal scales,and explored its structural basis.This dissertation mainly contains the following three aspects:(1)Research on methods of functional human brain parcellation.The functional units of the human brain are clearly hierarchical,performing specific cognitive and behavioral tasks through complex functional separation and integration.Dividing the functional units of the human brain at different spatiotemporal scales correctly is the basis for characterizing its information interaction patterns.However,the functional magnetic resonance imaging has a relatively low signal-to-noise ratio,which,as well as individual difference in brain structure,interferes with the division of brain functional area.Moreover,since the number of functional units of the brain at different spatiotemporal scales is unknown,how to select this number is a problem that has been plaguing the neuroscience community.In the second chapter,we proposed a new clustering algorithm,which attempted to automatically determine the number of brain functional units at a certain scale according to the distribution characteristics of the functional connection data.Thus,this proposed approach is noise-insensitive and independent of subjective parameter selection and prior knowledge.The effectiveness and robustness of the new clustering algorithm is verified through the functional parcellation experiments on the precuneus of the cerebral cortex,providing its superiority over the typical clustering algorithm.Meanwhile,multiple neurobiological markers indicate the validity of the results,our approach may be important for accurate brain parcellation.(2)Research on human brain functional parcellation and brain networks' topology at specific frequencies.We assume that the pattern of information interaction in the human brain varies across different time scales.Given this assumption,two problems are derived: whether the functional unit division is the same at different time scales,and whether the integration and separation modes between functional units are the same.In the third chapter,we used the global clustering and local gradient methods for cortical parcellation based on the coherence between the vertices of the cerebral cortex at different frequencies.We found that the definitions of brain networks and brain functional sub-areas varied at different frequencies,and so was the relationship between brain networks.Through the positioning and clustering of the hubs,we characterized the topology of the brain networks at different frequencies.Our research may provide a new perspective for the research on brain parcellation and information interaction among functional brain networks.(3)Research on gender identification based on three-dimensional cortical morphology.Using the grayscale features of the gray matter or white matter to calculate the concentration or volume information of a particular brain tissue and then performing multivariate pattern recognition has been a common practice.We believe that the measurement of brain tissue concentration is influenced by the morphology of the brain itself,while studies have shown that traditional cortical morphological features,such as cortical thickness and curvature,are insufficient to support individual-level pattern classification.And the classification accuracy based on those features is significantly lower than brain tissue-based classification,which does not effectively explain the relationship between brain tissue concentration and morphology.Given those facts,we proposed a hierarchical sparse representation classifier in the fourth chapter.Based on the three-dimensional morphological features of the cerebral cortex,we realized the gender recognition based on cortical morphology at the individual level for the first time,with an accuracy of 96.77%.Our results suggest that cortical morphology provides the most information on brain tissue concentration in magnetic resonance data,which provides new thoughts for the auxiliary diagnosis of brain diseases and the fusion of multi-center magnetic resonance imaging data.