The Predictive Value Of Renal Parenchymal Information For Renal Function Impairment In Patients With Adpkd

Objective1.The timely and early awareness of renal function impairment is of great clinical significance and value for autosomal dominant polycystic kidney disease(ADPKD),in which the role of total kidney volume(TKV)has been confirmed.On this basis,this study tried to demonstrate that renal parenchymal volume(RPV)may have a higher clinical value and guiding significance.2.We attempt to use radiomics to decode the more abundant intrinsic information of renal parenchyma,and construct a sensitive predictor of renal impairment using the combined model of RPV + intrinsic information to provide a new means for clinical monitoring of renal function and further improve the pre-hospital prevention and treatment of ADPKD patients.Methods1.A totals of 405 ADPKD patients with normal renal function were prospectively collected and followed-up for five years,with renal function tests and non-contrast computed tomography(CT)performed every six months.These patients with complete follow-up data were divided into a test set and a validation set according to the ratio of7:3,randomly.All CT images were segmented using the image segment editor module of the 3D-Slicer platform and the TKV and RPV were automatically calculated by the program attached to the segmentation software.The relationships between renal function impairment and TKV,RPV were explored using a multiple linear regression model and multiple logistic regression.2.A totals of 405 ADPKD patients with normal renal function were prospectively collected and followed-up for five years,with renal function tests and non-contrast CT performed every six months.These participants with complete follow-up data were randomly divided into a test set and a validation set(ratio=7:3).The segmentation operation was processed on the image processing module(Segment Editor)of the 3DSlicer platform and Py Radiomics modules based on Python3.6.8 was used for the extraction of radiomics features.A multiple linear regression model with covariates from clinical characteristics was constructed to investigate the relationship among radiomics features and e GFR.Multiple logistic regression was used to analyze the predictive value for renal function impairment and further combined them and RPV to comprehensively evaluate their predictive value for impairment.Results1.Excluding 65 subjects who were lost to follow-up,340 patients with ADPKD were eventually included in this study,in which the test set included 238 subjects,77 subjects were marked as the impairment group(IG group)and 161 were marked as the non-impairment group(NIG group),and the validation set included 102 subjects,29 subjects were marked as IG group and 73 were marked as the NIG group.Compared with the NIG group(baseline value),the IG group showed a higher proportion of PKD1 gene mutations,older age,more male patients,greater BMI index,a higher proportion of hypertension,higher Scr value and lower baseline e GFR(all p<0.05).2.Compared with the TKV,decreased RPV presented a closer relationship with renal function impairment,namely,with the impairment rate(RPV: 82.3% vs.TVK:67.1%)and e GFR(RPV: r=0.614,p<0.001 vs.TKV: r=-0.452,p<0.001).3.Compared with the TKV,the RPV showed higher predictive power for renal function impairment in ADPKD(RPV: AUC= 0.752[95%CI:0.692-0.805],p<0.001 vs.TKV: AUC=0.711[95%CI:0.649-0.768],p<0.001).4.A total of 851 radiomics features of renal parenchyma were extracted successfully,six of which were screened out through multistep feature selection(ICCsBoruta-collinearity).Four of these six features(Original-Long Run Emphasis,OriginalGray Level Non Uniformity,Wavelet-LHH-Complexity and Wavelet-LLL-Run Length Non Uniformity)showed significant differences between the groups(IG group vs.NIG group)(all p<0.001).5.The Original-Long Run Emphasis and Wavelet-LHH-Complexity also showed significant correlations with the baseline e GFR(r=-0.547,p<0.001;r=-0.516,p<0.001).According to further logistic regression analysis,two radiomics features presented satisfactory predictive performances(Original-Long Run Emphasis: AUC: 0.776;Wavelet-LHH-Complexit: AUC: 0.768;Original-Long Run Emphasis + Wavelet-LHHComplexity: AUC: 0.849).6.Combined prediction models of RPV and radiomics features showed a stronger predictive power(RPV + Original-Long Run Emphasis + Wavelet-LHH-Complexity,AUC=0.902;RPV + Original-Long Run Emphasis,AUC=0.848;RPV + Wavelet-LHHComplexity,AUC=0.856).7.The imaging biomarkers that were derived from radiomics were highly reproducible.The intra-and interobserver percentage differences were 0.74%±0.92%and 0.90%±1.03% for TKV;and 0.36%±0.72% and 0.56%±0.87% for RPV,respectively.Excellent intra-and interobserver reproducibility of important features were obtained: the range of intra-and interobserver ICCs were 0.932-0.955.In addition,the performance of the prediction model in the test set has also been satisfactorily verified in the validation set.Conclusion1.Compared with TKV,decreased RPV presented a closer relationship with renal function impairment,and showed a higher predictive power.Therefore,parenchymal volume may have more valuable clinical guiding significance for regular monitoring of renal function for patients with ADPKD.2.The more abundant intrinsic information of renal parenchyma could be decoded using radiomics analysis,and renal parenchyma information may be a sensitive biomarker of renal function impairment in ADPKD,which can provide a new approach for clinically monitoring renal function,and may greatly improve the pre-hospital prevention and treatment effects.