Effective Connectivity Analysis Of Multimodal Neuroimaging Data

As an important aspect in brain functional integration analysis,effective connectivity can reveal the information about "causes and effects" for the connectivity among neural activities.Functional magnetic resonance imaging(fMRI)and electroencephalogram(EEG)are two widely used non-invasive brain functional imaging techniques,which can respectively provide high spatial resolution and high temporal resolution data descriptions.This thesis focuses on the methodology and application research for the effective connectivity analysis of multimodal neuroimaging data.The main contents are as follows:Multivariate time-frequency Granger causality analysis of fMRI data.In this study,we proposed a combination method of time domain and frequency domain multivariate Granger causality analysis.There are two main advantages of this method:first,the surrogate data analysis is used to accomplish the group-level significance test of Granger causality measures that are of unknown statistical distributions.Second,the double verification from time domain as well as frequency domain analyses can improve the testing accuracy and eliminate the indirect Granger causal relationship,which is proved by simulation experiments.In terms of the application,we carried out research on abnormal patterns of effective connectivity among core neurocognitive networks in patients with idiopathic generalized epilepsy(IGE).IGE patients suffer long-term cognitive impairments,and present a higher incidence of psychosocial and psychiatric disturbances than healthy people.It is possible that the cognitive dysfunctions and higher psychopathological risk in IGE derive from disturbed causal relationship among brain networks.To test this hypothesis,we examined the effective connectivity across the salience network,default mode network,and central executive network using restingstate fMRI data collected from IGE patients and the matched healthy controls.Using the proposed method,we then observed significant differences in the effective connectivity graphs between groups(less connections for patients).Specifically,between-group statistical analysis revealed that relative to the healthy controls,the patients established significantly enhanced Granger causal influence from the dorsolateral prefrontal cortex to the dorsal anterior cingulate cortex,which is coherent both in the time and frequency domains analyses.Meanwhile,time domain analysis also revealed decreased Granger causal influence from the right fronto-insular cortex to the posterior cingulate cortex in the patients.These findings may provide new evidence for functional brain organization disruption underlying cognitive dysfunctions and psychopathological risk in IGE.Multimodal dynamic causal modelling and Bayesian fusion.This paper asks whether integrating multimodal EEG and fMRI data offers a better characterisation of functional brain architectures than either modality alone.First,based on the fact that “both EEG and BOLD measurements are observable consequences of the same neuronal activity”,we established a generative model of multimodal dynamic causal modelling(i.e.,multimodal DCM).We then on this basis proposed a multimodal fusion method,namely,Bayesian fusion.To be more specific,multimodal DCM generates both EEG and fMRI data from the same neuronal dynamics.Bayesian fusion means that we first obtain informative neuronal priors derived from DCM of EEG data,and then use them for subsequent DCM of fMRI data to further get haemodynamic parameter estimates.To illustrate this procedure,we generated multimodal EEG and fMRI dataset for a mismatch negativity paradigm,using biologically plausible model parameters.The results showed that Bayesian fusion(rather than single data modality)furnished better model parameter estimation,which was reflected by increases in free energy,a better match between real and predicted BOLD signals,and a reduction of posterior variance,indicating a shrinkage of uncertainty about model parameters.We quantified the benefits of multimodal fusion with the information gain pertaining to neuronal and haemodynamic parameters – as measured by the Kullback-Leibler divergence between their prior and posterior densities.Remarkably,this analysis suggested that EEG data can improve estimates of haemodynamic parameters,thereby furnishing proof-of-principle that Bayesian fusion of EEG and fMRI is necessary to resolve conditional dependencies between neuronal and haemodynamic estimators.These results suggest that Bayesian fusion may offer a useful approach that exploits the complementary temporal(EEG)and spatial(fMRI)precision of different data modalities.We envisage the procedure could be applied to any multimodal dataset that can be explained by a DCM with a common neuronal parameterisation.Spectral dynamic causal modelling of resting-state fMRI data.This work innovatively proposes an algorithm that combines functional connectivity analysis and effective connectivity analysis.First,functional connectivity analysis is used to confirm regions of interest(ROIs).Then,effective connectivity analysis is performed on the basis of the selected ROIs,in order to further explore of the directionality of these brain connections.Spectral DCM is used for effective connectivity analysis,which is now the predominant method for the DCM of resting-state fMRI analysis because of its high computational efficiency.Parametric empirical Bayes and Bayesian model reduction are finally used for the group-level analysis of effective connectivity parameters.The above algorithm provides a new idea about the selection of brain regions for the effective connectivity analysis of resting-state fMRI data,which is based upon the point that functional connectivity is the necessary non-sufficient condition for the effective connectivity.Subsequently,we applied this method to explore the functional interaction of supplementary motor area(SMA)and the whole brain for the taxi driver and the control groups.We found that changes in brain functional connectivity related to driving are specific to the pre-SMA,but not extended to the SMA proper.The pre-SMA showed enhanced functional connectivity with several prefrontal regions in the driver group.The latter effective connectivity analysis further revealed that the pre-SMA is located at a high level in the hierarchy,the connectivity from pre-SMA to other brain regions are inhibitory feedback connections.In addition,the effective connectivity from pre-SMA to the left superior frontal gyrus has the predictive validity for the group difference.These results indicate that specific skills acquired through training can be encoded by functionally connected brain structures in the resting-state.