Research On Characteristic Of Dynamic Brain Network In Patients With Mental Disease Based On Resting State FMRI

Resting-state functional magnetic resonance imaging(rs-fMRI)technology provides a new way to detect brain dysfunction,which plays an increasingly important role in the early intervention and diagnosis of patients with mental disease.We can perform brain function analysis from regions and networks of the brain to study the pathological characteristics of mental diseases.Based on the above background,the present research is divided into the following two parts:In the first part of this thesis,we used time-varying functional connectivity(FC)analysis to study the dynamic changes of brain activity in major depressive disorder(MDD)patients.The present work is the first study to investigate the state-dependent property(related to a specific connectivity state)of the abnormal dynamic FC in MDD patients.First,the fMRI data of 49 MDD patients and 54 healthy controls(HCs)were preprocessed,and 49 brain regions belonging to 7 brain functional subnetworks were identified using independent component analysis.Secondly,time-varying functional networks were constructed using sliding windows analysis and dynamic FC analysis,and K-means clustering was used to identify the sub-states of whole brain activity.In this study,weakly-connected state and strongly-connected state were obtained respectively.At last,the differences of functional connections between the two groups were calculated in weakly-connected state and strongly-connected state respectively.The dynamics of functional connections with significant differences were correlated with the severity of depressive symptoms in MDD patients.Compared with HCs,MDD patients showed increased mean dwell time and reduced FC between subnetworks and within subnetwork in the weakly-connected state.Dynamics of reduced FC between default mode network and cognitive control network as well as within cognitive control network predicted individual differences in depression symptom severity.The findings provide a new perspective for understanding the state-dependent neurophysiological mechanisms of MDD.The second part of this thesis explored the dynamic information flow mechanism in the default mode network of autism spectrum disorder(ASD)patients using dynamic directed effective connectivity(EC)analysis,and evaluated the effectiveness of dynamic direct EC and its time domain characteristics in distinguishing ASD from HCs.First,the fMRI data of 35 ASD patients and 62 HCs were preprocessed,and 10 brain regions in the default mode network were extracted using independent component analysis.Then,we used Granger causality analysis based on the sliding window method to construct the dynamic directional EC networks,which were used to explore the flow of information from one brain area to another and the strength change of this causal effect in DMN of ASD in the time-varying analysis.K-means clustering was used to identify the sub-states of directional EC in the default mode network.In this study,the strongly-EC state and weakly-EC state were obtained respectively.The differences between groups of directional effective connections were further calculated in the strongly EC state and weakly EC state respectively.The relationships were explored between dynamics of directional EC with significant differences and clinical scales of ASD patients.Compared with HCs,the ASD showed significantly decreased mean dwell time in the strongly-EC state.In both States,the abnormal directional effective connections within DMN were found in ASD patients.The results of the correlation analysis demonstrated that the dynamics of abnormal directional EC in DMN were related to the ASD patient's cognitive impairment(abnormality of social function and communication,and repetitive stereotyped behavior),confirming the key role of DMN dysfunction in the neurophysiological mechanism of ASD.Finally,this work adopted ten-fold cross-validation to evaluate classification performance and used the observed abnormal directional EC in each state and its time domain characteristics to verify its effectiveness in ASD recognition.We found that when the two are combined as classification features,the accuracy of ASD identification can reach 81.55%.The results showed that the dynamic directed EC characteristics are helpful for the auxiliary diagnosis of ASD.