Research On P300 Large-scale Network Based On EEG And FMRI Multi-modality Information Fusion

P300 is associated with a variety of cognitive functions,such as attention,intelligence,and working memory.The realization of these functions depends on the interaction between large-scale brain regions.Although studies relying on functional magnetic resonance imaging(fMRI)or electroencephalography(EEG)have performed some analysis on the P300 signal sources and the corresponding brain networks,there are few studies related to synchronous EEG-fMRI,as well as reports corresponding to the P300 large-scale brain network.In fact,relevant studies have reported large P300 differences between individuals,but the neural mechanisms behind the differences are still not very clear.Therefore,based on synchronous EEG-fMRI,this paper carries out related research on P300 large-scale brain network,which mainly includes the following three aspects:1.P300 large-scale network analysis based on spatial brain map.With the Power264 brain template and the canonical correlation analysis(CCA)method,we construct corresponding large-scale functional networks of rest and task fMRI states.We find that in the large-scale network corresponding to the rest state,the default network(DMN)interacts strongly with multiple sub-networks such as the visual network,fronto-parietal task control(FPN)and salience network.This kind of good interaction is the basis for the completion of brain function.When switching from the rest state to the run1,the multiple origin regions of the P300 are activated,which makes the participation of the FPN stronger,so connection of FPN-Salience is stronger.In run2,in response to the demand for continuously performing complex tasks,the core node of the DMN is weakened,and the interaction between the cingulo-opercular task control network(CON)and the salience network is strengthened again,which enhances the information processing in multiple stages related to P300 induction.2.P300 large-scale network analysis based on space-time decomposition.We use independent component analysis(ICA)to perform spatio-temporal decomposition of fMRI signals,and obtain nodes of large-scale networks and corresponding time processes.Group ICA separates 11 sub-networks related to the P300 task,such as extrastriate/primary visual network(Extra/Prim VN),dorsal attention network(DAN),left/right frontal-parietal network(L/RFPN)and somato-motor network(SMN).Based on these sub-networks,we construct a P300 large-scale functional network.At the same time,the analysis of the synchronized EEG data find that the target stimulus induce a clear P300 waveform.We find that the P300 amplitude of the subjects is significantly negatively correlated with the clustering coefficient,global efficiency,and local efficiency of the P300 large-scale network,and is significantly positively correlated with the length of the characteristic path.This means that individuals with higher amplitudes only need to consume less network resources when performing tasks.Individuals with higher P300 amplitudes own more efficient brains in the rest state,that is,the "starting point" of the task state is higher.Therefore,the high-amplitude individuals in the task state do not need to consume too much network resources.3.Dynamic large-scale network analysis of P300 based on sliding time window.We segment time series of independent components by sliding the time window to build P300 functional networks.The results of cluster analysis show that under the optimal window length,the P300 dynamic large-scale network can be summarized into two different modes,namely,the strong functional network connection responding to the target stimulus and the weaker functional network connection responding to the standard stimulus.In the large-scale network corresponding to the target stimulus,the RFPN exhibits stronger laterality;and it is necessary to focus attention in the field of view when performing the identification task,which causes the Extra VN to participate more strongly.As compensation,LFPN and Prim VN have strong participation in large-scale networks corresponding to standard stimulus.The different strategies of the above response targets and standard stimuli also explain that the target stimulus and the standard stimulus induce different signals(the target stimulus induces a clear P300 waveform).Therefore,we can obtain more information about the large-scale functional network corresponding to the brain's response to different stimuli from the time series of the P300 task state.