Gray-White Matter Structural And Neurovascular Coupling Impairment In Patients With End-stage Renal Disease Undergoing Hemodialysis:a Multimodal MRI Study

Objective:To explore the changes of gray matter volume and cortical thickness,the atrophy patterns of subcortical structures and hippocampal subfields,the abnormalities of white matter microstructure and the impairment of neurovascular coupling(NVC)function in patients with end-stage renal disease undergoing hemodialysis(HD)using multimodal MRI imaging.Association of these structural and functional changes with laboratory indicators,cerebral small vessel disease(CSVD)and cognitive function were evaluated to clarify the impact of gray-white matter structural damage and NVC dysfunction on cognitive in HD patients and to identify related clinical risk factors,with a view to providing imaging markers and potential interventions for the diagnosis and treatment of cognitive impairment in HD patients.Materials and Methods:(1)HD patients and healthy volunteers without cognitive decline(healthy control,HC)were prospectively enrolled.Demographic data,basic clinical data(dialysis duration,single-pool kinetic transfer/volume urea measurements[sp Kt/V],etc.),laboratory indicators(hemoglobin[Hb],creatinine[Cre],uric acid(UA),serum urea[Urea],total cholesterol,triglyceride[TG],homocysteine[HCY],albumin,blood sodium,blood calcium,blood phosphorus,corrected calcium[c Ca],parathyroid hormone[PTH]and C-reactive protein[CRP],etc.)and neuropsychological scale(Montreal cognitive assessment[Mo CA],Rey’s auditory verbal learning test[RAVLT],trail making test part A[TMT-A],trail making test part B[TMT-B]and digital symbol substitution test[DSST])were recorded.Traditional MRI sequence,3D-T1WI,diffusion tensor imaging(DTI),3D pseudo‐continuous ASL(p CASL)and rest-state function MRI(rs-f MRI)images were acquired on a Philips Ingenia CX 3.0 T MRI.The CSVD MRI features were assessed,including periventricular hyperintense(PVH),deep white matter hyperintense(DWMH),enlarged perivascular space in centrum semiovale(CS-EPVS)and basal ganglia(BG-EPVS),lacune,cerebral microbleeds(CMBs)and the total burden of CSVD.United Imaging artificial intelligence analysis software was used to automatically segment the white matter hyperintensity(WMH)to calculate the volume and proportion.(2)Thirty-nine HD patients and 51 HC were prospectively enrolled.Statistical parametric mapping(SPM 12)and Free Surfer 7.2.0 software were used for VBM and SBM analysis and the differences in gray matter volume and cortex thickness between the two groups were compared.Partial correlation or multiple linear regression was used to analyze the association between clinical data,laboratory indicators,CSVD and cognitive function and gray matter volume and cortical thickness in HD patients.P<0.05 indicates a statistically significant difference.(3)Thirty-nine HD patients and 51 HC were prospectively enrolled.The subcortical shape was evaluated using Shape analysis at the vertex level.Free Surfer7.2.0 software was applied to perform hippocampal subfields segmentation.The volumes of subcortical structures and hippocampal subfields between HD and HC were compared using covariance analysis.Group differences in subcortical shape were assessed using“Randomize”in FSL and threshold-free cluster enhancement was applied to perform multiple comparison corrections.In addition,ROC curve analysis was performed to evaluate the diagnostic performance of subcortical shape and hippocampal subfields volume in discriminating HD from HC.Partial correlation or multiple linear regression was used to analyze the association between clinical data,laboratory indicators,CSVD and cognitive function and subcortical structural abnormalities in HD patients.The significant threshold was set at P<0.05.(4)Thirty-eight HD patients and 43 HC were prospectively enrolled.17 fiber bundles were successfully identified based on automatic fiber quantification(AFQ).The DTI metrics of each participant at 100 equidistant nodes along each fiber tract were quantified.A non-parametric permutation-based statistical analysis(5,000permutations),with FWE procedure for multiple comparisons was utilized and P<0.05was considered statistically significant.The“random forest”function in R software(R4.2.2)was used to construct an RF classification model to determine the white matter fiber that can best predict disease diagnosis.Partial correlation or multiple linear regression was used to analyze the association between clinical data,laboratory indicators,CSVD and cognitive function and white matter microstructures in HD patients.The significant threshold was set at P<0.05.(5)Thirty-three HD patients and 35 HC were prospectively enrolled.SPM12software was used to process the ASL images to obtain cerebral blood flow(CBF)maps.DPARSF toolkit was used to preprocess rs-f MRI data to obtain amplitudes of low-frequency fluctuation(ALFF),fractional ALFF(f ALFF),regional homogeneity(Re Ho),functional connectivity density(FCD)and functional connectivity strengths(FCS)maps.Global and regional NVC patterns based on the whole gray matter template and the AAL 90 template were obtained,including CBF-ALFF、CBF-f ALFF、CBF-Re Ho、CBF-FCD and CBF-FCS.Covariance analysis was applied to compare the NVC alterations between HD and HC groups.Partial correlation or multiple linear regression was used to analyze the association between clinical data,laboratory indicators,CSVD and cognitive function and NVC alterations in HD patients.P<0.05was considered statistically significant.Results:(1)Compared with the HC group,the volume of gray matter in the HD group was significantly reduced,including bilateral superior cerebellar peduncle,left temporal pole:superior temporal gyrus,left putamen,right calcarine fissure and surrounding cortex,bilateral lingual gyrus,left heschl gyrus,right cuneus and bilateral postcentral gyrus(P<0.05,voxel-wise FWE corrected,voxel size>100 mm3).The cortical thickness in the left superior frontal,left supramarginal,left parsopercularis,bilateral postcentral,left lateral orbitofrontal,bilateral superior temporal,left lingual,bilateral lateral occipital,right caudal middle frontal,right medial orbitofrontal,right rostral middle frontal,right isthmus cingulate,right inferior parietal and right cuneus was reduced in the HD group(P<0.05,FDR corrected,voxel size>100 mm~2).Partial correlation showed that PTH was negatively correlated with the volume of the left heschl gyrus(r=-0.576,P=0.001)and the right cuneus(r=-0.563,P=0.001),c Ca was positively correlated with the right postcentral gyrus volume(r=0.407,P=0.029)and the cortical thickness of the left lingual(r=0.413,P=0.023),dialysis duration was negatively correlated with the cortical thickness of the right medial orbitofrontal(r=-0.419,P=0.021),Cre had a negative correlation with the cortical thickness of the right isthmus cingulate(r=-0.413,P=0.023)and Hb had a positive correlation with the cortical thickness of the right cuneus(r=0.442,P=0.015).Multiple linear regression analysis showed that the increase of WMH grades,EPVS score and total burden of CSVD were independent factors for the decrease of gray matter volume and cortical thickness in some brain regions(P<0.05).The volume of left lingual gyrus(β=0.441,P=0.009)and the cortical thickness of the left superior temporal(β=0.360,P=0.041)were independent factors of attention score of Mo CA,the cortical thickness of the right medial orbitofrontal(β=0.305,P=0.041)and the right postcentral(β=0.590,P<0.001)was independent factor of abstraction score and orientation score of Mo CA,respectively.The cortical thickness of the left superior temporal(β=0.342,P=0.022)and right postcentral(β=0.292,P=0.026)were independent influencing factors of Trial 1-5 in RAVLT,the cortical thickness of the inferior parietal was an independent influencing factor of delayed recall(β=0.258,P=0.024)and recognition of RAVLT(β=0.396,P=0.009).(2)Subcortical volumetric analysis showed that the volume of bilateral accumbens,bilateral thalamus and left amygdala were significantly reduced in the HD group(P<0.003,FWE corrected).Shape analysis showed that compared with the HC group,HD patients had surface atrophy in multiple subcortical structures,including brainstem,bilateral accumbens,bilateral amygdala,bilateral caudate,bilateral thalamus,left hippocampus,bilateral putamen and right pallidus(P<0.05,TFCE corrected).ROC curve analysis demonstrated that the classifications based on the shape alteration of right amygdala resulted in higher performance(AUC=0.773).Hippocampal subfields volumetric analysis showed that HD patients had reduced volumes in the whole hippocampus and subfields,including bilateral hippocampal tail,bilateral subiculum head,bilateral molecular layer body,bilateral fimbria,bilateral HATA,right CA1 head,right molecular layer head,right GC-DG body,right whole hippocampus body and right whole hippocampus head(P<0.05,FDR corrected).ROC curve analysis demonstrated that the classifications based on the left hippocampal tail volume resulted in higher performance(AUC=0.713).Partial correlation showed that the mean surface distances(MSD)of the left putamen was positively correlated with sp Kt/V(r=0.420,P=0.023)and negatively correlated with TG(r=-0.486,P=0.008),PTH was negatively correlated with MSD of the left thalamus(r=-0.452,P=0.014).Hb was positively correlated with the volume of GC-DG body(r=0.419,P=0.027).Multiple linear regression analysis showed that the total burden score of CSVD,the grade and proportion of WMH were independent factors of MSD values of bilateral caudate and bilateral thalamus(P<0.05).MSD value of the left thalamus was an independent factor of recognition score of RAVLT(β=0.531,P=0.025).The volume of the right hippocampal tail was an independent of abstraction score of Mo CA(β=0.358,P=0.024),the volume of the left fimbria was an independent influencing factor of Trial 2 in RAVLT(β=0.295,P=0.043)and DSST(β=0.273,P=0.030).(3)Compared with the HC group,HD patients exhibited widespread damaged tract profiles.Pointwise level analysis showed that HD patients had significantly lower FA and higher MD,Ax D and RD values in specific fiber segments of thalamic radiation(ATR),corticospinal(CST),callosum forceps major,callosum forceps minor,inferior fronto-occipital fasciculus(IFOF),superior longitudinal fasciculus(SLF)and uncinate(UF)(P<0.05,FWE corrected).RF classification results showed that the pointwise level of MD and tract level of RD profile had better discrimination ability in distinguishing between HD and HC groups.Based on the pointwise level of MD,the analysis of the variable importance showed that the two important variables contributing to HD vs.HC classification were left SLF(nodes 1-100)and right IFOF(nodes 1-46).Based on the tract level of RD profile,the analysis of the variable importance showed that the two important variables contributing to HD vs.HC classification were bilateral IFOF.Partial correlation showed that Urea was positively correlated with MD value of left SLF(r=0.391,P=0.036),sp Kt/V was negatively correlated with the mean MD value of right IFOF(nodes 1-46)(r=-0.403,P=0.030),UA had a positive correlation with mean RD of overall right IFOF(r=0.380,P=0.042)and right IFOF(nodes 1-55)(r=0.449,P=0.015).Multiple linear regression analysis showed that the total burden of CSVD was an independent factor of FA value of callosum forceps minor(nodes 9-28)(β=-0.333,P=0.032),MD value of the overall right IFOF(β=0.362,P=0.023)and RD value of the overall right IFOF(β=0.337,P=0.033).PVH was an independent influencing factor of Ax D value of ATR(nodes 1-28).FA value of callosum forceps minor(nodes 9-28)was the independent influencing factor of TMT-B(β=-0.386,P=0.013)and DSST(β=0.340,P=0.005).(4)Compared with HC,HD patients showed significantly decreased CBF-ALFF,CBF-f ALFF,CBF-Re Ho,CBF-FCD and CBF-FCS coupling at the whole brain gray matter level(P<0.01,FWE corrected).Decreased CBF-ALFF coefficient was found in 5 brain regions of the AAL90 Atlas,including bilateral median cingulate and paracingulate gyri(DCG),right middle occipital gyrus(MOG),left supramarginal gyrus(SMG)and left precuneus(PCUN)in HD patients.Increased CBF-ALFF coefficient was found in bilateral caudate nucleus(CAU)in HD patients.In CBF-f ALFF coefficient,significant differences were found in 5 brain regions of the AAL90 Atlas in HD patients,including right cuneus(CUN),left SMG,bilateral superior temporal gyrus(STG)and left middle temporal gyrus(MTG).Decreased CBF-Re Ho coefficient was found in 7 brain regions of the AAL90 Atlas,including bilateral calcarine fissure and surrounding cortex(CAL),right inferior occipital gyrus(IOG),left STG,bilateral MTG and right inferior temporal gyrus(ITG)in HD patients.Increased CBF-Re Ho coefficient was found in right CAU in HD patients.CBF-FCD and CBF-FCS in HD group were lower than HC group in bilateral CAL.There were widespread correlations between NVC and laboratory indicators,CSVD and cognitive scores(P<0.05)Conclusion:(1)HD patients have extensive cortical and subcortical structures atrophy,decreased gray matter volume,cortical thickness thinning,subcortical structures and hippocampal subfields atrophy are closely related to cognitive decline,mainly manifested in attention,memory and executive function.(2)Patients with HD have extensive,hemispherically symmetrical abnormalities of white matter microstructures.The fragility of white matter microstructural integrity may only occur in specific fiber segments and is strongly associated with cognitive decline,mainly manifested in executive function,visual memory and attention.(3)Patients with HD exhibit widespread NVC dysfunction in brain regions associated with visual network,DMN and ventral attention network,mainly manifested in the decline of cognitive domain functions such as attention,abstraction and memory.(4)Increased CSVD burden,anemia,dyslipidemia,inadequate dialysis,and accumulation of UTs are the influencing factors of cerebral cortex and subcortical structure atrophy,white matter microstructural vulnerability and NVC dysfunction in HD patients.