Research On Characteristics Of Brain Network Via Naturalistic Functional MRI

Current non-invasive magnetic resonance technology(MRI)enables researchers to inspect the human brain effectively and efficiently.Most encouragingly,functional magnetic resonance imaging(f MRI)technology makes measuring and recording functional activity of brain possible.This provides new strategy and clue for uncovering the mystery of the brain architecture and its functional characteristic.In addition,the representation of functional brain network is the focus of research on structure and characteristics of brain function.Current f MRI studies are mainly based on traditional task design or resting state.However,task paradigms usually adopt highly simplified and artificial stimuli,which are not ecologically relevant to natural real-life experience.Therefore,neuroscientists have proposed a novel neuroimaging paradigms – naturalistic stimuli,which largely restore most of the activities in everyday life,and can be applied to the research of perception,cognition and language or easily applied to patient populations.However,the dynamic and unconstrained nature of naturalistic stimuli poses big challenges for precisely identifying neural activities during various cognitive processes.Therefore,it is necessary to propose new methods for naturalistic f MRI studies.To address these concerns,in this thesis,sparse coding,inter-subject functional correlation and dynamic causal modelling are employed to investigate the characteristics of brain network under natural stimuli.The main contents and contributions of this thesis are summarized as follows:1.Functional brain network analysis based on sparse representation algorithm.? Resting-state brain network analysis and its clinical application to desease diagnosis based on sparse representation algorithm.While sparse coding has been widely used in the field of computer vision for a long time,few studies have applied it to the studies of functional magnetic resonance imaging.Neuroscience studies show that sparsity is the characteristic of neuronal activity,and the sparse combinations of different sets of neurons can accomplish different functions.To explore whether neural activity at the functional magnetic resonance image level shows the same sparsity characteristics,and whether sparse coding can identify functional networks with highly complex and resolution,we thus apply sparse representation to the whole-brain resting state f MRI data and apply it to patient populations.The results show that sparse coding can identify meaningful resting state brain networks.? Holistic atlas of naturalistic brain networks analysis based on sparse representation algorithm.Based on the inherent characteristic of naturalistic stimuli that the same naturalistic stimulus can evoke consistent brain activities across different subjects,we propose a group-wise sparse coding method,which aims to identify consistent brain function networks among different subjects.Our study reveal that group-wise sparse coding method can decompose the whole-brain naturalistic f MRI signals into many meaningful brain networks,which have correspondence between subjects and groups.To further investigate the reliability of brain networks and the influence of over-completion and different sparse coding parameters,we then conducted repeated experiments session,and evaluated test-retest reliability of each brain functional network generated by different sparse coding parameters,and the effects on sparse representation caused by over-completion problem.In addition,we compare the group-wise sparse coding with several independent component analysis-based methods,which demonsrated the superiority of our method in terms of the definition of brain functional networks,the reliability of brain functional network.2.The extensions and applications of naturalistic brain networks derived by sparse representation.? Group difference analysis under emotional cognition based on naturalistic brain networks.As group-wise sparse coding can identify consistent functional networks across subjects,we then use this method to investigate the group differences of brain functional networks.Therefore,we employ group-wise sparse coding to naturalistic f MRI signals to investigate the gender differences under emotional cognition.Our results show that sparse representation can effectively learn group differences in emotion process-related functional networks.? Application of whole-brain naturalistic brain networks in disease diagnosis.Since group-wise sparse coding can be used to explore group differences in healthy populations,we then investigate whether it can benefit clinical studies.We thus apply group-wise sparse representation to a variety of naturalistic f MRI data of cocaine dependent and pathological gambling patients,suggesting that sparse representation can define the differences between patient and healthy groups in functional networks.3.Functional architecture of brain networks under naturalistic stimuli.? The hierarchical architecture of brain function under naturalistic stimuli.Many neuroscience studies reveal that the sensory information of the external world is first entered into primary sensory cortex,then into the secondary sensory cortex,and finally into the higher-order cortex.However,for a long time,few studies have specifically defined the hierarchical architecture of brain.The traditional method inter-subject correlation is a commonly-used method to define the brain regions with consistent neural activities across subjects under naturalistic stimuli,which is usually used for characterizing the functional architecture of brain,but the binary division of brain derived from this method can oversimplify the hierarchy of brain function.Based on this method,we propose a novel method,Inter-subject Functional correlation,which can define the functional interactions between different brain regions across subjects,to define the extrinsic-intrinsic system of brain.Our results reveal the extrinsic and intrinsic systems of that are responsible for processing the extrinsic and intrinsic information respectively,and bring novel sight of the hierarchical architecture of brain function.? Inter-subject functional variability analysis based on functional hierarchy of brain under naturalistic stimuli.Based on the defined extrinsic-intrinsic tendency of each brain,we further explore the relationship between this tendency of each brain region and inter-subject functional variability.Our results demonstrate that the intrinsic system is specifically related to individual experience.This study is meaningful for identifying the functional organization of brain under naturalistic condition.4.Functional brain networks analysis based on improved naturalistic paradigm.? Comparison between improved naturalistic paradigm and traditional task-based design in brain functional network analysis.Due to the dynamic and unconstrained nature of traditional naturalistic paradigm,model-driven approaches are usually not suitable for naturalistic dataset.To address this problem,we design an improved naturalistic paradigm by combining the traditional naturalistic paradigm with task design.We then compare the improved naturalistic paradigm with task design in terms of functional brain network analysis.? Effective connectivity analysis based on improved naturalistic design.Based on the improved naturalistic design,we adopt dynamic causal modelling to investigate effective connectivity between brain regions,and analyze the correlation between the effective connectivity and behavioral data.