Research On Large-scale Brain Networks For Motor Imagery Based On FMRI

As a high-level cognitive process,motor imagery is normally defined as the mental rehearsal of specific actions from a first-person perspective without physical movement.Many studies have reported that MI can be widely used in physical function rehabilitation,sports training and brain-computer interface(BCI).But there are always dissension in the theoretical foundation and intrinsic neural mechanism of MI.Hence,the further research of the neural mechanism underlying MI will facilitate the clarification of its execution mode and provide the necessary neural basis for practical applications.This dissertation mainly investigates the network patterns of MI by using the brain network analysis at large-scale level.The current work focuses on the large-scale patterns of functional networks during resting-state and two MI tasks,and the large-scale dynamic patterns of functional networks during MI tasks.This paper mainly includes two parts:1.Research on the large-scale patterns of functional networks during resting-state and left/right-hand MI.Based on canonical correlation analysis(CCA)and brain atlas,we explored a new method to construct the large-scale brain networks.The simulation experiment found that the CCA can effectively find the coupling relationship between different brain regions,and reveal the essential large-scale network patterns.From the large-scale network patterns constructed by CCA for MI fMRI,we find that there are certain differences existed between resting-state and MI.In resting-state,it shows strong interactions between Default mode network(DMN)and other brain networks.This reflects that the brain maintains the basic network structure for processing input tasks.Compared with the resting-state,the edges of DMN connecting with other networks are decreased.Meanwhile,the edges of Fronto-parietal Task Control network(FPN)and Salience network(SN)are significantly increased.The connection strength of these two networks are significantly correlated with the MI-BCI performance,which shows that FPN and SN plays an important role in MI.2.Research on the large-scale dynamic patterns of functional networks during left/right-hand MI.Based on independent component analysis(ICA),sliding window and clustering,we constructed the large-scale dynamic patterns of functional networks during MI,and probed the differences between left-hand MI and right-hand MI by network-based statistic(NBS)approach.The findings show that there are four inherent states in the large-scale dynamic network patterns during MI.Through these states,we found that DAN and DMN are mutually antagonistic,and DAN and DMN consistently support the cognitive activity during MI.At the same time,ECN is continuously active to ensure that participants can deal with MI tasks.In addition,the results of NBS showed that there exists certain degree of similarity in large-scale dynamic network patterns between left-hand MI and right-hand MI.By adjusting the connections of the dynamic networks slightly,the switch of network can be completed,and DAN plays the role of the central node in the task switching.In summary,based on f MRI data,the canonical correlation analysis,independent component analysis,sliding window and cluster analysis methods were used to explore the static and dynamic connectivity patterns of large-scale brain network during motor imagery in this paper from two different levels.Our results demonstrated that the validity of the canonical correlation analysis and revealed that the default mode network,fronto-parietal network,and salient network play an important role in left/right-hand motor imagery.Meanwhile,our results showed that the large-scale dynamic network of motor imagery tasks contained four intrinsic states,and the dynamic interaction of the executive control network,the dorsal attention network,and the sensory-motion network regulated the execution of the motor imagery task.Therefore,this study will promote us understanding of the underlying neural mechanisms of motor imagery at the large-scale network level.