Research On Multi-Scale Brain Network

Human brain might be the most exquisite and complex system in the uni-verse.Trying to explore and understand the brain’s mechanisms is probably the most important subject for us human beings.As a new tool,brain network anal-ysis has played an increasingly important role in illuminating human cognition,its variation over development and aging,and its alteration in disease or injury.Human brain is also a system which could be analyzed at spatial scales ranging from that of individual cells to brain regions and with temporal precision rang-ing from sub-millisecond to that of the entire lifespan.Research and analysis on brain network at various scales could offer quite different perspectives on under-standing brain structure and function.Therefore,this dissertation dedicated to exploring brain network from both macro-scale and micro-scale,and mainly in-cluded:at macro-scale level,we constructed functional brain networks based on resting-state fMRI data and then conducted quantitative study on human brain lifespan development;at micro-scale level,we constructed a spatiotemporal firing model of neuron-network from a new point of view and proved the convergence of the firing model under given conditions.Chapter 1 is the introduction part.In this chapter,firstly,we introduced the background and significance of this dissertation.Secondly,we briefly expounded the neuronal structure,the neurological mechanism of action potential,the struc-ture and function of synapses,and the mathematical models of neurons.Thirdly,we described the basic principle of MRI,different kinds of brain networks and their research situations,topological attributes of complex network,and appli-cations of machine learning in brain networks.Finally,arrangements of contents and chapters were given.In Chapter 2,we made an exploration on human brain lifespan development based on resting-state fMRI data.On one hand,we explored the lifespan changes of human brain from three perspectives:at the edge-level or subnetwork-level,we employed linear and quadratic regression models to probe the human brain lifespan development process;from the perspective of time-frequency analysis,we explored the lifespan changes of fALFF of voxels;principal component analysis was performed on the FC matrices across all the subjects to explore meaningful components and we acquired components which was especially correlated with age.On the other hand,coefficients across the components,edge features after a newly proposed feature reduction method as well as temporal features based on fALFF,were extracted as predictor variables and three different regression models were learned to make prediction of brain age.In this stage,we employed two open source data sets,one data set for training the models and the other for validation.We adopted the method of K-fold cross validation on the training data set to choose the best model.The predictive models extracted sufficient information to provide comparatively accurate predictions of brain age.The predictions could also be regarded as a potential biomarker of senescence or maturity.In Chapter 3,we constructed a spatiotemporal firing model of neuron-network from a new point of view in order to depict the dynamic property of neuronal networks.Moreover,we introduced the concept of Brownian Web which is a family of coalescing Brownian motions starting from every point in space and time R × R from the field of stochastic process.We showed that under diffusive scaling this model converges in distribution to the nonnegative Brownian Web.Chapter 4 is summary and prospect part.We concluded the main work of this dissertation and came up with the possible research direction in the near future based on the existing problems.