Experimental And Clinical Study On Evaluating Renal Ischemia Reperfusion Injury By MRI Texture Analysis

Part I Experimental study on evaluating renal ischemia-reperfusion injury by MRI texture analysisObjective:To explore the value of MRI texture analysis in evaluating early renal ischemia-reperfusion injury(IRI)and to analyze the correlation between MRI texture features and the histopathological features by establishing renal IRI model in rabbits.Material and Methods:Sixty-two healthy New Zealand rabbits were randomly divided into two groups,renal ischemia reperfusion(IR)group(n=54)and the sham group(n=8).Rabbits in the IR group underwent the left kidney surgery by clamping the left renal artery for 60 mins and then releasing the clamp to establish renal IRI model.The sham group underwent the same operation as the IR group,but without clamping the left renal artery.The rabbits in the IR group and the sham group underwent conventional MRI,susceptibility weighted imaging(SWI)and blood-oxygen level dependent magnetic resonance imaging(BOLD MRI)examination before the operations and at 1 h,12 h,24 h,and 48 h after the operations.Immediately after completion of the MRI examination,8 rabbits in the IR group were randomly sacrificed at the different time points(subgroup pre-IR,IR 1 h,IR 12 h,IR 24 h,and IR 48 h).The other fourteen rabbits in the IR group(subgroup IR-L)and the eight rabbits in the sham group were sacrificed after MRI examination at 48 h after the operations.The left kidney was harvested for histopathological examination.The single item score of the histopathological features,the microvessel density,and the pathological total score were calculated.The texture features of the left kidney based on T2WI,SWI,and BOLD were extracted using MaZda software.Multiple dimensionality reduction for MRI texture features were performed,and the texture features based on T2WI,SWI,and BOLD associated with the pathological total score were selected.Differences in the single item score of the histopathological features,the microvessel density,and the pathological total score among the subgroups with different time points were determined using the non-parametric Kruskal-Wallis test followed by pairwaise comparisons.The relationship between the texture features based on T2WI,SWI,and BOLD with the single item score of the histopathological features,the microvessel density(MVD),or the pathological total score was assessed by Pearson correlation coefficients.Repeated measures analysis was used to compare the differences in the texture features of the IR-L group and the sham group based on T2WI,SWI,and BOLD among the different time points.The diagnostic performance of the texture features based on T2WI,SWI,and BOLD in differentiating the different IRI time points(pre-IR vs IR 1h+12 h,pre-IR vs IR 24 h+48 h,and IR 1h+12 h vs IR 24 h+48 h)was assessed by receiver operating characteristic(ROC)curve analysis,and z test was performed to compare the difference between the areas under the ROC curves.P<0.05 was considered to indicate a significant difference in all statistical tests.Results:There were significant differences in renal tubular brush border destruction,renal tubular epithelial edema,necrosis,cast,interstitial inflammatory cell infiltration,MVD,and the pathological total score among pre-IR group,IR 1 h group,IR 12 h group,IR 24 h group,and IR 48 h group(all P<0.0001).After Multiple dimensionality reduction,four texture features based on T2WI(T2WI_S(3,-3)InvDfMom,T2WI_S(0,5)DifEntrp,T2WI_Horzl_GLevNonU,and T2WI_GrMean),three texture features based on SWI(SWI_S(3,-3)Contrast,SWI_S(4,4)DifEntrp,and SWI_S(4,-4)Correlat),and three texture features based on BOLD(BOLD_Perc.99%,BOLD_S(0,3)DifVarnc,and BOLD_S(4,-4)Correlat)were selected.T2WI_S(3,-3)InvDfMom has the highest correlation with renal tubular brush border destruction,renal tubular epithelial edema,necrosis,cast(r=0.56,P<0.0001;r=-0.58,P<0.0001;r=0.62,P<0.0001;r=0.69,P<0.0001).BOLD_S(4,-4)Correlat has the highest correlation with interstitial inflammatory cell infiltration(r=0.63,P<0.0001).SWI_S(4,4)DifEntrp has the highest correlation with renal tubular cast(r=0.61,P<0.0001).There were no significant differences in T2WI_S(3,-3)InvDfMom?SWI_S(4,4)DifEntrp?BOLD_S(4,-4)Correlat in the sham group among the different time point of the operations(all P>0.05).During the 48 h period after the operations,T2WI_S(3,-3)InvDfMom and BOLD_S(4,-4)Correlat in the IR-L group gradully increased and SWI_S(4,4)DifEntrp gradully decreased.The AUC for T2WI_S(3,-3)InvDfMom?SWI_S(4,4)DifEntrp?BOLD_S(4,-4)Correlat in differentiating pre-IR with IR 1 h+12 h was 0.762,0.722,and 0.699,and there were no significant differences in the AUCs among T2WI_S(3,-3)InvDfMom?SWI_S(4,4)DifEntrp?BOLD_S(4,-4)Correlat(all P>0.05).The AUC for T2WI_S(3,-3)InvDfMom?SWI_S(4,4)DifEntrp?BOLD_S(4,-4)Correlat in differentiating pre-IR with IR 24 h+48 h was 0.842,0.813,and 0.690,and the AUC for T2WI_S(3,-3)InvDfMom was significantly higher than that for BOLD_S(4,-4)Correlat(z=1.965,P=0.049).The AUC for T2WI_S(3,-3)InvDfMom in differentiating IR 1 h+12 h with IR 24 h+48 h was 0.662,with 67.05%accuracy,79.55%sensitivity,and 54.55%specificity.Conclusion:MRI texture analysis can assess the dynamic changes of the early renal IRI at different time points and the correlation with histopathology,and can provide valuable information for evaluating early renal IRI.Part II Clinical study on evaluating renal allograft injury by MRI texture analysisObjective:To explore the value of MRI texture analysis in evaluating renal allograft injury by extracting renal allograft MRI texture features associated with estimated glomerular filtration rate(eGFR).Material and Methods:Sixty-six patients who underwent allograft renal transplantation(42 male,24 female;age range,22-63 years;median age,40 years)between November 2013 and December 2016 were included in this study.The fasting intravenous blood samples for all the patients were collected in the morning before MRI examination.Serum creatinine and serum cystatin C were measured,and estimated glomerular filtration rate(eGFR)was calculated according to the equation of serum creatinine and serum cystatin C published by American chronic kidney disease epidemiology collaboration.All the patients were divided into three groups according to their eGFR:normal renal allograft function(nRAF)group(eGFR? 60 ml/min/1.73 m2,n=15),mild to moderate renal allograft injury(mRAI)group(30 ?eGFR<60 ml/min/1.73 m2,n=18),and severe renal allograft injury(sRAI)group(eGFR<30 ml/min/1.73 m2,n=33).All the patients underwent conventional MRI,susceptibility weighted imaging(SWI),and blood-oxygen level dependent magnetic resonance imaging(BOLD MRI)examination.The texture features of renal allograft based on T2WI,SWI,and BOLD were extracted using MaZda software.Multiple dimensionality reduction for MRI texture features were performed,and the texture features based on T2WI,SWI,and BOLD with absolute correlation coefficient of eGFR greater than or equal to 0.3(P<0.05)and also with the highest Z value for Boruta algorithm were selected.Differences in serum creatinine,serum cystatin C and eGFR among the nRAF group,the mRAI group and the sRAI group were compared using one-way analysis of variance(ANOVA)followed by LSD test.The diagnostic performance of the selected texture features based on T2WI,SWI,and BOLD in differentiating the nRAF group with the mRAI group or the sRAI group and in differentiating the mRAI group with the sRAI group was assessed by receiver operating characteristic(ROC)curve analysis,and z test was performed to compare the difference between the areas under the ROC curves.P<0.05 was considered to indicate a significant difference in all statistical tests.Results:Serum creatinine and serum cystatin C in the sRAI group were significantly higher than those in the nRAF group and in the mRAI group(all P<0.0001).After Multiple dimensionality reduction,one texture features based on T2WI(T2WI_Perc.50%),one texture features based on SWI(SWI_Perc.01%),and two texture features based on BOLD(BOLD_S(4,4)Contrast,BOLD_S(5,5)Correlat)with absolute correlation coefficient of eGFR greater than or equal to 0.3(P<0.05)and also with the highest Z value for Boruta algorithm were selected.There were significant differences in T2WI_Perc.50%,SWI_Perc.01%,BOLD_S(4,4)Contrast,and BOLD_S(5,5)Correlat between the nRAF group and the sRAI group(all P<0.05).T2WI_Perc.50%in the mRAI group was significantly lower than that in the nRAF group(P=0.001).BOLD_S(4,4)Contrast in the sRAI group was significantly lower than that in the mRAI group(P=0.032).The AUC for T2WI_Perc.50%,SWI_Perc.01%,and BOLD_S(5,5)Correlat in differentiating the nRAF group with the mRAI group was 0.785,0.720,and 0.700,and there were no significant differences in the AUCs among T2WI_Perc.50%,SWI_Perc.01%,and BOLD_S(5,5)Correlat(all P>0.05).The AUC for T2WI_Perc.50%,SWI_Perc.01%,BOLD_S(4,4)Contrast,and BOLD_S(5,5)Correlat in differentiating the nRAF group with the sRAI group was 0.687,0.733,0.784,and 0.737,and there were no significant differences in the AUCs among T2WI_Perc.50%,SWI_Perc.01%,BOLD_S(4,4)Contrast,and BOLD_S(5,5)Correlat(all P>0.05).The AUC for BOLD_S(4,4)Contrast in differentiating the mRAI group with the sRAI group was 0.667,with 66.67%accuracy,54.55%sensitivity,and 88.89%specificity.Conclusion:There is a correlation between MRI texture features and renal allograft function.MRI texture analysis can provide valuable information for evaluating renal allograft injury and can be used as a non-invasive tool for monitoring renal allograft function.