The Application Value Of Multiparametric Magnetic Resonance-based Radiomics In The Diagnosis Of Prostate Cancer

Aim:To develop and validate a radiomics model based on prostate multiparametric MRI(mpMRI)with prostate-specific antigen(PSA)in the diagnostic gray zone to reduce unnecessary biopsies.Methods:Patients with suspected prostate cancer whose PSA was between 4-10 ng/mL were retrospectively included from December 2015 to March 2018.All patients underwent mpMRI examination before biopsy.Patients were assigned to the training set and validation set according to the 2:1 ratio.The region of interest was outlined in each horizontal T2 weighted imaging,diffusion weighted imaging and dynamic contrast enhancement(DCE)imaging.The lesions were segmentated and features were extracted.A clinical-imaging model is constructed from clinical information.A random forest classifier is used to construct each radiomics model.And a combination model is constructed from the radiomics model and the clinical-imaging model.The area under the receiver operating characteristic curve of each model were calculated.The performance of the model is shown by sensitivity,specificity,negative predictive value and positive predictive value.Results:The combined model(AUC in the primary and validation cohorts:0.956[95%CI:0.951-0.961]and 0.933[95%CI:0.918-0.948],respectively)performed better than the clinic-radiological model(AUC in the primary and validation cohorts:0.806[95%CI:0.793-0.819]and 0.858[95%CI:0.835-0.881],respectively).The combined model yielded NPV and specificity of 0.824 and 0.737,respectively.The clinic-radiological model achieved negative predictive value(NPV)and specificity of 0.875 and 0.921,respectively.Decision curve analysis showed that the combined model was more beneficial than the clinic-radiological model in predicting PCa.Conclusions:We developed and validated an mpMRI-based radiomics model to improve the NPV and specificity of PCa using PSA in the gray zone.This model might provide improved clinical information to reduce unnecessary biopsies in prostate cancer.Aim:Biopsy Gleason score(GS)is crucial for prostate cancer(PCa)treatment decision making.Upgrading in GS from biopsy to radical prostatectomy(RP)puts a proportion of patients at risk of undertreatment.To develop and validate a radiomics model based on multiparametric magnetic resonance imaging(mpMRI)to predict upgrading of PCa from biopsy to RP to avoid undertreatment.Methods:One hundred and sixty-six RP-confirmed PCa patients(training cohort,n=116;validation cohort,n=50)were retrospectively included.A total of 4404 features were extracted from T2-weighted imaging(T2WI),diffusion weighted imaging(DWI)derived apparent diffusion coefficient(ADC)and dynamic contrast enhancement(DCE)sequences.We applied Mutual Information Maximization criterion to identify the optimal features on each sequence.Multivariate logistic regression analysis was used to develop predictive models and a radiomics nomogram.Student' s t or Chi-square was used to assess the differences in clinicopathologic data between training and validation cohorts.Receiver operating characteristic(ROC)curve analysis was performed and area under the curve(AUC)was calculated for training and validation cohorts.The performance of predictive models and nomogram was evaluated with respect to discrimination,calibration and clinical usefulnessResults:The radiomics model incorporating signatures from the three sequences achieved better performance than any single sequence,the combined model incorporating radiomics signature and selected clinical features(AUC values of 0.914 and 0.910 in the training and validation cohorts)outperformed the clinical model(AUC values of 0.677 and 0.646 in the training and validation cohorts)and radiomics model(AUC values of 0.899 and 0.868 in the training and validation cohorts).Decision curve analysis showed the radiomics nomogram was clinically useful.Conclusion:Radiomics based on mpMRI has potential to predict upgrading of prostate cancer from biopsy to RP.Aim:The clinically recommended treatments for low-risk prostate cancer(PCa)and high-risk PCa are different.Differential diagnosis of the these before the biopsy can guide clinical decisions more personally and benefit them in the management of PCa.Therefore,an radiomics model based on prostate multi-parameter MRI(mpMRI)was developed and validated to risk stratification to differentiate between low-risk and high-risk PCa to guide clinical decisions.Methods:A retrospective analysis of 183 patients(main population and validation population=6:4)received mpMRI of the prostate before pathology after radical prostatectomy(RP).The radiomic features were extracted from the regions of interest(ROI),which segmented from each T2 weighted,fat suppressed T2 weighted,diffusion weighted and dynamic contrast enhanced(DCE)image.Features are selected using maximum correlated minimum redundancy(mRMR)and lasso regression(Lasso)algorithms.After single-factor and multi-factor analysis,the clinical biomarkers were selected to construct a clinical model;the support vector machine was used to establish the radiomics model in the training set;the radiomics model and clinical biomarkers were used as multivariate logistic regression input variables to construct a fusion model.The performance of the model was verified in the training and validation sets,including the area under the ROC curve(AUC),accuracy,sensitivity,specificity,positive predictive value(PPV),and negative predictive value(NPV).Results:In this study,prostate volume and PSAD were used to construct a clinical model.The AUC in the training set was 0.8(95%CI:0.71-0.88)and the AUC in the validation set was 0.75(95%CI:0.61-0.88).The AUC of the radiomics model on the training set is 0.84(95%CI:0.77-0.92),and the AUC of the validation set is 0.79(95%CI:0.68-0.90).The radiomics model,prostate volume,and PSAD were used as multivariate logistic regression input variables to construct a fusion model.The AUC in the training set was 0.89(95%CI:0.82-0.95),and the AUC in the validation set was 0.83(95%CI:0.73-0.93).Compared with clinical models(training set sensitivity 0.756 and NPV 0.587;validation set:sensitivity 0.609 and NPV 0.429,perspectively),the use of radiomics models can improve the sensitivity of PCa risk stratification(training set:0.879 and Validation set:0.905,perspectively)and NPV(training set:0.927 and validation set;0.938,perspectively);as well as compared to the clinical model(training set:specificity 0.818 and PPV 0.908;validation set:specificity 0.714 and NPV 0.838,perspectively),Using the fusion model can increase the specificity of PCa risk stratification(training set:0.939 and validation set:0.952,perspectively)and PPV(training set:0.963 and validation set:0.971,perspectively).Conclusion:We constructed and validated the mpMRI-based radiomics model for risk stratification of PCa,which can improve the detection rate of low-risk PCa and high-risk PCa,and develop personalized treatment plans for different patients who benefit from PCa management.