Radiomic nomogram based on bi-parametric magnetic resonance imaging to predict the International Society of Urological Pathology grading ≥ 3 prostate cancer: a multicenter study

Abstract

Purpose

To create a reliable radiomic nomogram for the prediction of the International Society of Urological Pathology (ISUP) grading ≥ 3 PCa patients.

Methods

395 patients with verified PCa were obtained from three different hospitals. The patients were divided into training, internal validation, and two external validation groups. A radiomic signature (rad-score) extracted from T2WI, DWI, and apparent diffusion coefficient (ADC) maps was constructed in the training cohort. Eight clinical features were performed to develop a clinical model using univariate, and multivariate logistic regression. The combined model incorporated the radiomic signature and clinical model. The model's performance was assessed by receiver operating characteristic (ROC) curve.

Results

Rad-score, MRI T-stage, and ADC value were significant predictors of ISUP ≥ 3 PCa. A nomogram of these three factors was shown to have greater diagnostic accuracy than using only the radiomic signature or clinical model alone. The area under ROC curve was 0.85, 0.88, 0.81, 0.81 for the training, internal, and two external validation cohorts, respectively. In the stratified analysis based on MR scanners model, the AUC of predicting ISUP ≥ 3 PCa for GE, Siemens, and combined groups were 0.84, 0.83 and 0.84, respectively, in the combined training group and internal validation group.

Conclusions

The proposed nomogram has the potential to predict the differentiation degree of ISUP PCa patients.