| ObjectiveAmong the numerous medical complications of type 2 diabetes Mellitus(T2DM),cognitive impairment has been regarded as a major public health problem due to its huge socioeconomic cost and lack of effective treatment.But the underlying neural mechanisms underlying cognitive impairment in T2DM have not been elucidated.With the development and application of functional magnetic resonance(fMRI)technology,the dynamic functional network connectivity analysis approach is of great value in the study of neuropsychiatric diseases for it can capture dynamic changes in brain function.In addition,the syndromedifferentiation of six meridians can fully reflect the syndrome type characteristics as well as the transmission and change patterns of T2DM,which may include the occurrence and progression information of T2DM cognitive impairment.Therefore,this study was divided into two parts.Part 1:We used dynamic functional network connectivity analysis to investigate the difference of brain functional connectivity between T2DM and HC,as well as its correlation with cognitive function,aimed to provide clues for the early diagnosis of cognitive impairment in T2DM brain imaging and improve the prognosis of patients.Part 2:To investigate whether there are differences in cognitive function among T2DM patients with different syndrome types under the syndrome-differentiation of six meridians system,and to evaluate the syndrome susceptibility of T2DM cognitive impairment.Meanwhile combined the brain functional imaging data and clinical data based on dynamic functional network connectivity analysis method,aimed to provide insights for exploring the underling mechanism of T2DM cognitive impairment,as well as to provide a scientific basis for the theory of "Prevention of disease occurrence and Prevention of disease progression" in traditional Chinese medicine for the diagnosis and treatment of T2DM cognitive impairment.MethodsPart 11.Participants:All T2DM patients were recruited through the outpatient and inpatient department of Endocrinology of the First Affiliated Hospital of Guangzhou University of Chinese Medicine,and all healthy controls were recruited through advertisements in the surrounding communities of Guangzhou University of Chinese Medicine.After strict inclusion and exclusion criteria screening,56 T2DM patients and 46 HC were included.After image preprocessing,2 of T2DM patients were removed for the excessive head movement,resulting 54 T2DM patients and 46 HC finally included.2.Cognitive function assessment:Before imaging data acquisition,Chinese version of the Montreal cognitive assessment scale test(MoCA),trail making test(TMT),grooved pegboard test(Gpb)and digit symbol substitution test(DSST)were performed on each participant to assess their cognitive function.3.Image acquisition:A 3.0 Tesla Siemens Prisma MRI scanner with a 64-channel head coil was used for MRI data acquisition.4.Imaging preprocessing:Imaging preprocessing was performed on GRETNA package.Imaging data were converted from DICOM into NIfTI form firstly,then the initial 10 time points of each participant’s rs-fMRI data were discarded to eliminate the destabilizing effects which may generated by the maladjustment of machine and participants at the beginning of the scan,leaving 950 images for further analysis.Registered T1 images to mean volume,spatially normalized the functional images onto the standard Montreal Neurological Institute(MNI)space with warping parameters estimated from coregistered T1 images,resliced the voxel size as 3 × 3 × 3 mm3 using a Diffeomorphic Anatomical Registration Through Exponentiated Lie algebra(DARTEL)strategy,and then smoothed the normalized data with a 6 mm full width half max(FWHM)Gaussian kernel to make an increase of the signal-tonoise ratio.Data would be excluded if the displacement or head rotation parameters exceeded 2 mm or 2° in this study.After data preprocessing,two of the 102 participants were excluded because of the excessive head motion.All the above steps were repeated for the MRI data from the remaining 100 subjects owing to that DARTEL normalization based on the registration to all subjects.No individual was excluded in this repetition process.5.Independent component analysis and brain functional network screening:The rs-fMRI data were decomposed by subject-level principal component analysis(PCA)for dimension reduction which totally generated 60 components,then these individual data were conducted temporal concatenation and reduced by a group-level PCA which finally yielded 40 independent components.The infomax algorithm was applied to extract the independent spatial map and time course of each component.The ICASSO algorithm was repeated 20 times to improve the stability of the decomposition.The subject-specific spatial maps and time courses were back-reconstructed from group-level independent components by a group independent component analysis(GICA)approach.Combined with two templates of functional network atlases and the independent component analysis(ICA)spatial multiple regression values and the spatial overlapped minimal with ventricular and edge regions of the brain.31 of the 40 obtained independent components(ICs)were selected as physiologically relevant Intrinsic Connectivity Networks(ICNs)and finally were classified into 12 functional networks.6.Dynamic brain functional network connectivity analysis:Firstly,the linear detrend number was set up as 3 and time courses were processed by 3DDESPIKE to ensure that artifactual spikes did not interfere the signal analysis,then a high frequency cutoff of 0.15 Hz was performed as low-pass filtering for all time courses of the 31 selected ICNs.Secondly,a sliding window approach was applied to calculate dynamic FNC.A window size of 44s(88TR)was selected according to previous studies,as it has been reported that a window length between the range 30-60s could effectively capture dynamic information The window was shifted with a sliding step size of 1 TR(=0.5 s).A tapered window was created by convolving a Gaussian(σ=3)with a rectangular function.For each window,a full correlation matrix was calculated across 31 ICNs.Resulting in a total of 862 dynamic functional network connectivity matrices.Thirdly,k-means clustering method was employed on all slidingwindow FNC matrixes of all subjects to detect specific functional network connectivity patterns.The squared Euclidean distance algorithm(500 iterates and 150 replicates)was used to estimate the similarity between different time windows in the analysis.Based on the elbow criterion(ratio of within-to between-cluster distances),the optimal number of distinguishable whole-brain functional network connectivity pattern reoccurrence clusters were estimated to be four(k=4).Cluster centroids were then used as starting points to cluster all the dynamic functional network connectivity windows from all subjects into four functional network connectivity states.Finally,computed the dFNC indices for each participant.7.Statistical analyses:A Shapiro-Wilk test was initially used to verify Gaussianity of continuous variables,then two-sample t-test was used to evaluate differences in group comparisons if the data were conformed to normal distribution,nonparametric MannWhitney U test was used otherwise.Chi-square test was used on categorical variables.Pearson correlation analysis was used for correlation analysis among all values in line with normal distribution,otherwise Spearman correlation analysis was used.For the correlation analysis of interferential variables,partial correlation analysis was used to control the variables.P<0.05 was considered statistically significant.Part 21.Participants:All T2DM individuals were recruited from the outpatient and inpatient departments of endocrinology of the First Affiliated Hospital of Guangzhou University of Chinese Medicine.After strict inclusion and exclusion criteria screening,24 Shaoyin yang deficiency complicated with cold dampness syndrome T2DM patients and 39 failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome T2DM patients.2.Cognitive function assessment:Before imaging data acquisition,Chinese version of MoCA,TMT,Gpb and DSST were performed on each participant to assess their cognitive function.3.Image acquisition:A 3.0 Tesla Siemens Prisma MRI scanner with a 64-channel head coil was used for MRI data acquisition.4.Imaging preprocessing:Imaging preprocessing was performed on GRETNA package.Imaging data were converted from DICOM into NIfTI form firstly,then the initial 10 time points of each participant’s rs-fMRI data were discarded to eliminate the destabilizing effects.Registered T1 images to mean volume,spatially normalized the functional images onto the standard MNI space with warping parameters estimated from coregistered T1 images,resliced the voxel size as 3 × 3 × 3 mm3 using DARTEL strategy,and then smoothed the normalized data with a 6 mm FWHM Gaussian kernel to make an increase of the signal-tonoise ratio.Data would be excluded if the displacement or head rotation parameters exceeded 2 mm or 2° in this study.No individual was excluded in this process.5.Independent component analysis and brain functional network screening:The rs-fMRI data were decomposed by subject-level PCA for dimension reduction which totally generated 63 components,then these individual data were conducted temporal concatenation and reduced by a group-level PCA which finally yielded 41 independent components.The infomax algorithm was applied to extract the independent spatial map and time course of each component.The ICASSO algorithm was repeated 20 times to improve the stability of the decomposition.The subject-specific spatial maps and time courses were backreconstructed from group-level independent components by a GICA approach.Combined with two templates of functional network atlases and the ICA spatial multiple regression values and the spatial overlapped minimal with ventricular and edge regions of the brain.32 of the 41 obtained ICs were selected as physiologically relevant ICNs and finally were classified into 12 functional networks.6.Dynamic brain functional network connectivity analysis:Dynamic brain functional network connectivity analysis:Firstly,the linear detrend number was set up as 3 and time courses were processed by 3DDESPIKE to ensure that artifactual spikes did not interfere the signal analysis,then a high frequency cutoff of 0.15 Hz was performed as low-pass filtering for all time courses of the 32 selected ICNs.Secondly,a sliding window approach was applied to calculate dynamic FNC.A window size of 44s(88TR)was selected according to previous studies,as it has been reported that a window length between the range 30-60s could effectively capture dynamic information The window was shifted with a sliding step size of 1 TR(=0.5 s).A tapered window was created by convolving a Gaussian(σ=3)with a rectangular function.For each window,a full correlation matrix was calculated across 32 ICNs.Resulting in a total of 862 dynamic functional network connectivity matrices.Thirdly,k-means clustering method was employed on all sliding-window FNC matrixes of all subjects to detect specific functional network connectivity patterns.The squared Euclidean distance algorithm(500 iterates and 150 replicates)was used to estimate the similarity between different time windows in the analysis.Cluster centroids were then used as starting points to cluster all the dynamic functional network connectivity windows from all subjects into four functional network connectivity states.Finally,computed the dFNC indices for each participant.7.Statistical analyses:A Shapiro-Wilk test was initially used to verify Gaussianity of continuous variables,then two-sample t-test was used to evaluate differences in group comparisons if the data were conformed to normal distribution,nonparametric MannWhitney U test was used otherwise.Chi-square test was used on categorical variables.Pearson correlation analysis was used for correlation analysis among all values in line with normal distribution,otherwise Spearman correlation analysis was used.For the correlation analysis of interferential variables,partial correlation analysis was used to control the variables.P<0.05 was considered statistically significant.ResultsPart 11.54 T2DM patients and 46 HC subjects were finally included.There were no statistically significant differences in gender,age,education,systolic blood pressure level and diastolic blood pressure level between the two groups.The MoCA scores of HC group were significantly higher than that of T2DM group,there was no statistically difference in other cognitive function assessment indicators.2.A total of 31 independent components were screened from the 40 components obtained by group level independent component analysis and then were assigned to 12 functional networks.By using the sliding time window and K-means clustering analysis,the functional network matrices were clustered into 4 states.State 1 was characterized as a‘sparsely connected state’,state 3 was characterized as a‘highly segregated state’,and state 4 was characterized as a‘strong connectivity state’.3.We found in state 2,the connectivity strength between Right Executive Control Network(RECN)and Auditory Network(AUN)/Visual Network(VN)/Precuneus Network(PCUN)/Salience Network(SN)/Ventral Attention Network(VAN),and between PCUN and VAN was significant higher,but the connectivity strength between PCUN and Sensorimotor Network(SMN)and Cerebellar Network(CB)was significant lower in the T2DM group.In state 3,higher connectivity between RECN and VN/PCUN,while prominent lower connectivity between PCUN and SMN/CB in T2DM group.State 4 displayed higher connectivity between VN and Dorsal Attention Network(DAN)in T2DM group than HC group.4.Compared to the HC group,T2DM group possessed significantly decreased number of participants,fraction time(FT)and mean dwell time(MDT)in state 3,meanwhile remarkably increased FT and MDT in state 4.TMT_A completion time were mild negatively correlated with FT at state 1;Systolic blood pressure level were mild negatively correlated with the FT and MDT at state 3.MoCA scores were mild negatively correlated with the FT and MDT at state 4.Part 21.24 Shaoyin yang deficiency complicated with cold dampness syndrome T2DM patients and 39 failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome T2DM patients were finally included.There were no statistically significant differences in gender,age,education,systolic blood pressure level,diastolic blood pressure level,HbA1C fasting blood-glucose,fasting insulin,Gpb and DSST between the two groups.The MoCA scores of Shaoyin yang deficiency complicated with cold dampness syndrome group were significantly lower than that of failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome group,the TMT completion time of Shaoyin yang deficiency complicated with cold dampness syndrome group were significantly longer than that of failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome group.2.A total of 32 independent components were screened from the 41 components obtained by group level independent component analysis and then were assigned to 12 functional networks.By using the sliding time window and K-means clustering analysis,the functional network matrices were clustered into 4 states.State 1 was characterized as a‘strong connectivity state’ and state 3 was characterized as a‘sparsely connected state’.3.We found in state 1,the connectivity strength between RECN and VN was significant lower in Shaoyin yang deficiency complicated with cold dampness syndrome group.In state 3,lower connectivity between Ventral Default Mode Network(vDMN)and VN,while higher connectivity between Dorsal Default Mode Network(dDMN)and DAN in Shaoyin yang deficiency complicated with cold dampness syndrome group.State 4 displayed higher connectivity between PCUN and RECN,as well as within PCUN in Shaoyin yang deficiency complicated with cold dampness syndrome group.4.Compared to the failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome group,the Shaoyin yang deficiency complicated with cold dampness syndrome group possessed significantly decreased FT and MDT in state 1,meanwhile remarkably increased FT and MDT in state 3.Age was mild negatively correlated with FT at state 1 and number of transitions;TMT completion time was mild negatively correlated with FT at state 3;TMT A completion time were mild positively correlated with the systolic blood pressure level.Gpb_L completion time were mild positively correlated with HbA1C level.ConclusionPart 1Compared to the HC group,T2DM group possessed significantly decreased number of participants,FT and MDT in‘highly segregated state’,moreover,FT and MDT at this state were both mild negatively correlated with systolic blood pressure level;meanwhile remarkably increased FT and MDT in‘strong connectivity state’,moreover,FT and MDT at this state were both mild negatively correlated with MoCA scores.These suggested that‘highly segregated state’and‘strong connectivity state’may be the key differentiators between T2DM and HC,which may be associated with cognitive function changes.Our study revealed functional connectivity alterations about three crucial cognitive related networks including DMN,TPN and SN in T2DM patients,especially the antagonistic relationship between TPN and DMN.Part 2There were significant differences in cognitive function between Shaoyin yang deficiency complicated with cold dampness syndrome group and the failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome group.Compared to the failure of Shaoyang’s pivot function resulting into accumulation of gallbladder fire syndrome group,Shaoyin yang deficiency complicated with cold dampness syndrome group possessed significantly increased FT and MDT in‘sparsely connected state’,meanwhile decreased FT and MDT in‘strong connectivity state’,which suggested these may be the key differentiators between the two groups.The connectivity strength of DMN subnetworks between the two groups showed different trends in the same state,suggesting the necessity of dividing DMN into subnetworks. |
PDF Full Text Request