| With the aging of the population,the impact of Alzheimer’s disease(AD)and other senile diseases on people is becoming increasingly prominent,and the progress of brain science is urgently needed to provide new solutions for AD prevention and treatment.The human brain is a complex dynamic system shaped by relatively fixed anatomical connections,in which spontaneous spatial dynamic recombination of neural activities through nonlinear interactions among various neural units,aims at flexibly reconfiguring the functional network to adapt to the changing external environment.The study of brain network spatial dynamics and its mechanism is of great significance for the early diagnosis and pathogenesis of AD.In recent years,researchers have made some achievements in characterizing the functional spatial dynamics of the brain from the perspective of time reconstruction by combining complex network theory.However,data-driven empirical research methods can find some statistically significant phenomena,which cannot explain neurologically or conduct systematic studies on the development,evolution,or dynamics of the brain’s nervous system from the neural mechanism.This dissertation uses a brain network dynamics modeling method based on theory and data,integrates multi-mode image data,establishes a bridge between anatomical structure and functional dynamics,and explores the neural mechanism of brain spatial dynamics.Therefore,this dissertation takes computational neuroscience and network neuroscience as the main research methods,and adopts the method of combining empirical research and theoretical research to carry out in-depth and systematic analysis of brain spatial dynamics.Moreover,these research methods are applied to AD multi-modal image data to explore the anomalies spatial dynamics of AD and their underlying neural mechanisms.To be specific,the main innovative work and research achievements of this dissertation include the following four parts:(1)Construct a dynamic multi-layer brain network,a data-driven method is used to propose a node index of network recombination based on phase coherence,and explore spatial dynamics of the brain.In view of the limitations of the traditional sliding time window method,combined with rs-f MRI technology and nonlinear phase analysis method,this research constructs a dynamic multilayer brain network to explore the spatial dynamics mediated by low-frequency phase coherence,and improve the temporal resolution and sensitivity of the network.On this basis,a phase-coherent brain state extraction method based on leading eigenvector dynamics is introduced to dynamically characterize different brain states.Furthermore,the multi-layer network community reconstruction algorithm is introduced to propose node indexes of phase coherent dynamic network,which improves the research framework of spatial reorganization mode at the global,subnet and node levels.Then,based on this framework,the abnormal functional spatial dynamics of AD are explored,and a large number of abnormal features related to cognitive scales are found,which verifies the effectiveness of this research framework in rs-f MRI signal analysis.(2)Improve dynamic modeling method of brain network based on individual cortical reconstruction to explore the neural mechanism of brain spatial dynamics.Based on the datadriven empirical research,this dissertation uses a brain network dynamics modeling method based on theory and data to study the physiological mechanism of complex brain functions.However,the research on brain network dynamics modeling is still in its infancy,and the established model still has many shortcomings in terms of accuracy,efficiency,biological characteristics included,and interpretation of functional spatial dynamic mechanisms.In view of the above problems,this thesis uses computer technology to strengthen research from three aspects: automatic extraction of features of high-quality empirical data,automatic model inversion algorithm,and appropriate node model according to application scenarios,and proposes a brain network dynamics modeling method based on individual cortical reconstruction.The fitting of simulated data and empirical data of the model has reached the international excellent level,which confirms the effectiveness of the modeling method.On this basis,based on the full-parameter spatial working mode of the model,this research analyzes how the functional spatial dynamics observed in the resting state arise from the interaction between anatomical connections and intrinsic neurodynamics.It is further explained that the brain network dynamics model is a bridge connecting cross-modal experimental data,connection theory and phenomena,and undertakes the task of explaining and verifying the working mechanism of the brain.(3)Propose a modeling method for lesion effects of different structures to explore abnormal neural mechanisms.Through the unremitting efforts of interdisciplinary researchers,a large number of AD imaging markers have been discovered,but their pathogenesis is still unclear.To explore the abnormal neural mechanism of brain disease from the causal level,this research proposes a brain network dynamics model modeling method with different structural lesion effects,and models the relationship between anatomical abnormalities and functional dynamic changes from two aspects: remote white matter connections and local electrophysiological circuit disturbances.This method can be used to perturbing the brain system without ethical constraints and predict the response of brain activity to different structural injuries,which makes up for the limitations of empirical research.The results of focal virtual lesion research show that the lesion effect shows regional dependence,which can be predicted by the cortical gradient information of the lesion node and the topology index of the structural network.The multifocal virtual lesions study guided by AD imaging markers found that different brain regions or white matter junction lesions have different effects on global metastable perturbation,which can provide a reference for the study of the pathogenesis of AD.(4)Propose a brain network dynamics model construction method based on cortical gradient constraints to improve the model’s fitting degree to empirical f MRI data.Brain network dynamics modeling aims to balance complexity,mining the most important functional features of the brain from complex phenomena,and achieving a balance between the fineness of model depiction,the complexity of the model and the purpose of modeling.Based on the above findings,cortical gradients may be involved in the spatial dynamics of brain function,as well as findings from cellular structures,cross-species discoveries,and other computational neuroscientists,this research hypothesizes that the mechanisms of functional diversity in different brain regions may be related to cortical gradient heterogeneous information.Through the introduction of cortical heterogeneity prior knowledge to improve the construction method of brain network dynamics model,it is found that the improved method proposed in this thesis significantly improves the fitting of simulation data and empirical data,and promotes the development of brain network dynamics modeling technology.In summary,this research aims at the limitations of the current brain network dynamics modeling methods and the actual needs of the research on the neural mechanism of AD abnormality.The research method deeply integrates brain imaging technology and computational neuroscience theory,improves the spatial dynamic analysis method of brain networks,explores the dynamic characteristics of AD abnormality from multiple angles,improves the accuracy of brain network dynamics modeling,and provides references for the research on the pathogenesis of AD.The research results obtained have certain theoretical value and application values.This research is supported by the National Natural Science Foundation of China(61873178),and the Key R & D Program of Shanxi Province(201803D421047). |
PDF Full Text Request