| Part 1 Diagnostic value of combining PI-RADS v2.1 with PSA-derived indicators in prostate cancerObjectiveTo investigate the diagnostic value of Prostate Imaging Reporting and Data System version2.1(PI-RADS v2.1)for clinically significant prostate cancer(CsPCa),combined with PI-RADS v2.1 and PSA-derived indicators to improve the predictive value of CsPCa positive biopsy results that guide clinical biopsy decision-making.Materials and MethodsA retrospective collection of 524 patients who underwent MRI and performed prostate biopsy at Shenzhen people’s hospital from November 2014 to November 2019,including 181 cases of PCa(139 CsPCa,42 clinically insignificant prostate cancer)and 343 cases of benign prostatic hyperplasia(BPH).All cases were scored by PI-RADS v2.1 and clinical data were collected,including age,PSA,fPSA/tPSA,PV and PSAD.Univariate analysis was performed to screen out statistically significant indicators.Logistic regression was used to analyze the predictive value of PI-RADS v2.1 combined clinical indicators for CsPCa.ResultsThe differences of PI-RADS,age,PSA,fPSA/tPSA,PV and PSAD between PCa group and BPH group,CsPCa group and non-CsPCa group were statistically significant.PI-RADS v2.1 and PSAD had the highest diagnostic value,with AUC of 0.927 and 0.854,respectively.Logistic regression analysis revealed that PI-RADS v2.1 and PSAD were independent predictors of CsPCa,and the AUC of the regression model is 0.934.CsPCa detection rates were low when PI-RADS≤2,or PI-RADS=3 and PSAD≤0.28 ng/mL/mL.In contrast,when PI-RADS≥4 and PSAD≥0.15 ng/mL/mL was associated with the highest CsPCa detection rate.ConclusionCombining the PI-RADS v2.1 and PSAD did improve the predictive performance of CsPCa and assist in clinical biopsy decision-making.Patients with PI-RADS≤2,or PI-RADS=3 and PSAD≤0.28 ng/mL/mL can watchful waiting to avoid unnecessary biopsy.Part 2 Predictive value of MRI radiomics combined with clinical features and PI-RADS v2.1 for peripheral high-risk prostate cancerObjectiveTo explore the predictive value of MRI radiomics combined with clinical features and PI-RADS v2.1 for models of Gleason score(GS)risk grading of peripheral zone prostate cancer(PCa)Materials and MethodsA retrospective collection of 133 patients with peripheral PCa,who underwent MRI in Shenzhen people’s hospital,including 65 cases of GS low-risk PCa with GS<8 and 68 cases of GS high-risk PCa with GS≥8.2284 radiomics features were extracted from T2WI and ADC imaging for each patient.Combining important features selected by LASSO with PI-RADS v2.1 and clinical features,which differences between GS low-risk group and GS high-risk group were statistically significant to construct four types of model.The predictive value and generalization ability of the models were verified using 10-fold cross validation,the mean of AUC after cross validation were used to evaluate the predictive value of each model for GS high-risk PCaResultsThe differences of PI-RADS v2.1 and PSA between GS low-risk PCa group and GS high-risk group were statistically significant(P<0.05),AUC for diagnosing GS high-risk PCa were 0.649 and 0.690,respectively.Nine features were selected by LASSO from 2284 features,all of which were extracted from T2WI,and each feature was significantly correlated with GS risk grading Nine features combined with PI-RADS v2.1 and PSA construct four types of model,which can better predict the GS risk grading of peripheral PCa.Among them,the random forest model had the best predictive value,AUC was 0.800ConclusionThe predictive value of the models to stratify GS of peripheral PCa based on MRI radiomics combined with clinical features and PI-RADS v2.1 were higher than using PSA or PI-RADS alone.It is expected to provide a basis noninvasively for the GS risk grading of peripheral PCa before surgery to assist clinical determining the treatment plan and further achieve the goal of precision medicine. |
PDF Full Text Request