Preliminary Study On The Value Of The Scoring Based On BpMRI And Radiomics Combined With Clinical Variables For Assessing Extracapsulare Extension In Prostate Cancer

Part Ⅰ The Comparison of Modified ESUR Score and Mehralivand Grade Based on bpMRI for Assessing Extracapsulare Extension in Prostate CancerPurpose To compare the value of the modified ESUR score and Mehralivand grade based on bi-parametric MRI(bpMRI)in the assessment of extracapsulare extension(ECE)in prostate cancer(PCa),and to explore the potential value of clinical variables in assessing ECE.Methods Date of 235 patients of PCa confirmed by postoperative pathology,who underwent MRI between March 2019 and March 2022 in our hospital were retrospectively enrolled in this study.The clinical information included each patient’s age,prostate specific antigen(PSA)concentration at admission,lesion location distribution,biopsy Gleason Group,and the Prostate Imaging Report and Data System version 2.1(PI-RADS v2.1)score given by the radiologists.Two radiologists(reader 1 and 2)assessed the ECE by the modified ESUR score and Mehralivand grade without obtaining the pathological results and the inter-and intra-observer correlation coefficient(ICC)were calculated.The receiver operating characteristic curve(ROC)and Delong test were used to evaluate the performance of the two scoring methods.Then,univariate and multivariate binary logistics regression analysis were used to screen out the independent risk factors from clinical variables,which were combined with the scores of reader 1 to establish combined models.And the assessment ability of the two combined models and the two scoring methods were compared.Results Both scoring methods showed good inter-and intra-observer stability(both ICC>0.75).The modified ESUR score achieved an AUC of 0.696(95%CI:0.633～0.754)in reader 1 and 0.691(95%CI:0.627～0.749)in reader 2,and Mehralivand grade achieved an AUCs were 0.746(95%CI:0.685～0.800)in reader 1 and 0.753(95%CI:0.693～0.807)in reader 2.In Delong test,the AUC of Mehralivand grade in reader 1 was higher than that of the modified ESUR score in reader 1 and 2[0.746(95%CI:0.685～0.800)vs 0.696(95%CI:0.633～0.754)and 0.691(95%Cl:0.627～0.749),both P<0.05].And the AUC of Mehralivand grade in reader 2 was also higher than that of the modified ESUR score in reader 1 and 2[0.753(95%CI:0.693～0.807)vs 0.696(95%CI:0.633～0.754)and 0.691(95%CI:0.627～0.749),both P<0.05].Then,PSA,PI-RADS v2.1 score,biopsy Gleason Group,modified ESUR score and Mehralivand grade were all determined as the independent risk factors of ECE and the combined models were constructed by the integration of the independent risk factors and two scores of reader 1.The AUCs of the combined model 1 based on the modified ESUR score and the combined model 2 based on Mehralivand grade were 0.826(95%CI:0.773～0.879)and 0.841(95%CI:0.790～0.892),respectively,which were both higher than that in the separate modified ESUR score[0.826(95%CI:0.773～0.879)and 0.841(95%CI:0.790～0.892)vs 0.696(95%CI:0.633～0.754),both P<0.001],and also higher than that in the separate Mehralivand grade[0.826(95%CI:0.773～0.879)and 0.841(95%CI:0.790～0.892)vs 0.746(95%CI:0.685～0.800),both P<0.05].Conclusion Based on bpMRI,the Mehralivand grade showed better diagnostic performance for assessing ECE than the modified ESUR score.The combination of scoring methods and clinical variables can be a potential tool in the assessment of ECE.Part Ⅱ A Preliminary Study on the Value of a Combined bpMRI-Based Radiomics-Clinical Model for Assessing Extracapsular Extension in Prostate CancerPurpose To evaluate the value of a combined model considering bpMRI-based radiomics and clinical variables for assessing ECE in PCa.Methods Date of 392 patients of PCa confirmed by postoperative pathology,who underwent MRI between January 2018 and March 2022 in our hospital were retrospectively enrolled in this study.The clinical information included each patient’s age,PSA concentration at admission,lesion location distribution,biopsy Gleason Group,and the PI-RADS v2.1 score given by the radiologists.All patients were randomly divided into the training and validation sets at a ratio of 7:3.Radiologists assessed the ECE of all lesions by the Mehralivand grade and delineated all lesions without obtaining the pathological results.Radiomics features were extracted from T2WI,DWI.and ADC images in each lesion.A radiomics model was built by the features normalization of Z-Score,the redundant features remotion of RFE algorithm,and the iterative training of LASSO classifer.Then,univariate and multivariate binary logistics regression analysis were used to screen out independent risk factors from clinical variables to construct clinical model.Moreover,the combined model was constructed by the integration of the independent risk factors and the radiomics model,and the nomogram for assessing ECE was constructed based on the combined model.Finally,Delong test and the decision curve analysis were used to compare the performance in the assessment of ECE among clinical model,Mehralivand grade,radiomics model and combined model.Results In the training set,the AUCs of the combined model and radiomics model were 0.897(95%CI:0.861～0.934)and 0.882(95%CI:0.844～0.921),respectively.Based on the validation set.the AUC of the combined model was comparable to that of the radiomics model[0.894(95%CI:0.837～0.950)vs 0.835(95%CI:0.763～0.908),P>0.05].In addition,the combined model had a higher sensitivity(90.7%vs 77.8%)and accuracy(81.4%vs 76.3%)compared with the radiomics model.On the other hand,the AUCs of the clinical model were 0.805(95%CI:0.753～0.850)and 0.749(95%CI:0.661～0.824)in the training set and validation set,respectively,which were lower than those in the combined model[in the training set:0.805(95%CI:0.753～0.850)vs 0.897(95%CI:0.861～0.934),P=0.050;in the validation set:0.749(95%CI:0.661～0.824)vs 0.894(95%CI:0.837～0.950),P=0.050;both P<0.05].The AUCs of the Mehralivand grade in radiologists were 0.746(95%CI:0.690～0.796)and 0.774(95%CI:0.688～0.846)in the training set and validation set,respectively,which were also lower than those in the combined model[in the training set:0.746(95%CI:0.690～0.796)vs 0.897(95%CI:0.861～0.934),P<0.001;in the validation set:0.774(95%CI:0.688-0.846)vs 0.894(95%CI:0.837～0.950),P=0.010;both P<0.05].At last,the decision curve analysis implied that the combined model could obtain the maximum net clinical benefits compared with the clinical model,Mehralivand grade and radiomics model.Conclusion The combined bpMRI-based radiomics-clinical model has satisfactory estimated value for ECE in PCa patients comparing with the clinical model,Mehralivand grade,and radiomics model,which could be conducive to the preoperative individualized and accurate diagnosis and treatment of PCa patients.