Purpose We sought to develop diagnostic models incorporating mpMRI examination to identify PCa (Gleason score≥3+3) and CSPCa (Gleason score≥3+4) to reduce overdiagnosis and overtreatment. Methods We retrospectively identified 784 patients according to inclusion criteria between 2016 and 2020. The cohort was split into a training cohort of 548 (70%) patients and a validation cohort of 236 (30%) patients. Age, PSA derivatives, prostate volume, and mpMRI parameters were assessed as predictors for PCa and CSPCa. The multivariable models based on clinical parameters were evaluated using area under the curve (AUC), calibration plots, and decision curve analysis (DCA). Results Univariate analysis showed that age, tPSA, PSAD, prostate volume, MRI-PCa, MRI-seminal vesicle invasion, and MRI-lymph node invasion were significant predictors for both PCa and CSPCa (each p ≤0.001). PSAD has the highest diagnostic accuracy in predicting PCa (AUC=0.79) and CSPCa (AUC=0.79). The multivariable models for PCa (AUC=0.92, 95% CI: 0.88–0.96) and CSPCa (AUC=0.95, 95% CI: 0.92–0.97) were significantly higher than the combination of derivatives for PSA ( p =0.041 and 0.009 for PCa and CSPCa, respectively) or mpMRI (each p <0.001) in diagnostic accuracy. And the multivariable models for PCa and CSPCa illustrated better calibration and substantial improvement in DCA at threshold above 10%, compared with PSA or mpMRI derivatives. The PCa model with a 30% cutoff or CSPCa model with a 20% cutoff could spare the number of biopsies by 53%, and avoid the number of benign biopsies over 80%, while keeping a 95% sensitivity for detecting CSPCa. Conclusion Our multivariable models could reduce unnecessary biopsy without comprising the ability to diagnose CSPCa. Further prospective validation is required.
Background Machine learning has many attractive theoretic properties, specifically, the ability to handle non predefined relations. Additionally, studies have validated the clinical utility of mpMRI for the detection and localization of CSPCa (Gleason score ≥ 3 + 4). In this study, we sought to develop and compare machine-learning models incorporating mpMRI parameters with traditional logistic regression analysis for prediction of PCa (Gleason score ≥ 3 + 3) and CSPCa on initial biopsy. Methods A total of 688 patients with no prior prostate cancer diagnosis and tPSA ≤ 50 ng/ml, who underwent mpMRI and prostate biopsy were included between 2016 and 2020. We used four supervised machine-learning algorithms in a hypothesis-free manner to build models to predict PCa and CSPCa. The machine-learning models were compared to the logistic regression analysis using AUC, calibration plot, and decision curve analysis. Results The artificial neural network (ANN), support vector machine (SVM), and random forest (RF) yielded similar diagnostic accuracy with logistic regression, while classification and regression tree (CART, AUC = 0.834 and 0.867) had significantly lower diagnostic accuracy than logistic regression (AUC = 0.894 and 0.917) in prediction of PCa and CSPCa (all P < 0.05). However, the CART illustrated best calibration for PCa (SSR = 0.027) and CSPCa (SSR = 0.033). The ANN, SVM, RF, and LR for PCa had higher net benefit than CART across the threshold probabilities above 5%, and the five models for CSPCa displayed similar net benefit across the threshold probabilities below 40%. The RF (53% and 57%, respectively) and SVM (52% and 55%, respectively) for PCa and CSPCa spared more unnecessary biopsies than logistic regression (35% and 47%, respectively) at 95% sensitivity for detection of CSPCa. Conclusion Machine-learning models (SVM and RF) yielded similar diagnostic accuracy and net benefit, while spared more biopsies at 95% sensitivity for detection of CSPCa, compared with logistic regression. However, no method achieved desired performance. All methods should continue to be explored and used in complementary ways.
PurposeThe clinical utility of multiparametric magnetic resonance imaging (mpMRI) for the detection and localization of prostate cancer (PCa) has been evaluated and validated. However, the implementation of mpMRI into the clinical practice remains some burden of cost and availability for patients and society. We aimed to predict the results of prostate mpMRI using the clinical parameters and multivariable model to reduce unnecessary mpMRI scans.MethodsWe retrospectively identified 784 men who underwent mpMRI scans and subsequent prostate biopsy between 2016 and 2020 according to the inclusion criterion. The cohort was split into a training cohort of 548 (70%) patients and a validation cohort of 236 (30%) patients. Clinical parameters including age, prostate-specific antigen (PSA) derivates, and prostate volume (PV) were assessed as the predictors of mpMRI results. The mpMRI results were divided into groups according to the reports: “negative”, “equivocal”, and “suspicious” for the presence of PCa.ResultsUnivariate analysis showed that the total PSA (tPSA), free PSA (fPSA), PV, and PSA density (PSAD) were significant predictors for suspicious mpMRI (P < 0.05). The PSAD (AUC = 0.77) and tPSA (AUC = 0.74) outperformed fPSA (AUC = 0.68) and PV (AUC = 0.62) in the prediction of the mpMRI results. The multivariate model (AUC = 0.80) had a similar diagnostic accuracy with PSAD (P = 0.108), while higher than tPSA (P = 0.024) in predicting the mpMRI results. The multivariate model illustrated a better calibration and substantial improvement in the decision curve analysis (DCA) at a threshold above 20%. Using the PSAD with a 0.13 ng/ml2 cut-off could spare the number of mpMRI scans by 20%, keeping a 90% sensitivity in the prediction of suspicious MRI-PCa and missing three (3/73, 4%) clinically significant PCa cases. At the same sensitivity level, the multivariate model with a 32% cut-off could spare the number of mpMRI scans by 27%, missing only one (1/73, 1%) clinically significant PCa case.ConclusionOur multivariate model could reduce the number of unnecessary mpMRI scans without comprising the diagnostic ability of clinically significant PCa. Further prospective validation is required.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.