Comparisons of three scoring systems based on biparametric magnetic resonance imaging for prediction of clinically significant prostate cancer

Abstract

In this study, we aimed to validate and compare three scoring systems based on biparametric magnetic resonance imaging (bpMRI) for the detection of clinically significant prostate cancer (csPCa) in biopsy-naïve patients. In this study, we included patients who underwent MRI examinations between January 2018 and December 2022, with MRI-targeted fusion biopsy (MRGB) as the reference standard. The MRI findings were categorized using three bpMRI-based scorings, in all of them the diffusion-weighted imaging (DWI) was the dominant sequence for peripheral zone (PZ) and T2-weighed imaging (T2WI) was the dominant sequence for transition zone (TZ). We also used the Prostate Imaging Reporting and Data System version (PI-RADS) v2.1 to evaluate each lesion. For each scoring, we calculated the sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and area under the receiver operating characteristic (ROC) curves (AUC). The calculated AUC for three bpMRI-based scorings were 83.2% (95% CI 78.8%–87.6%), 85.0% (95% CI 80.8%–89.3%), 82.9% (95% CI 78.4%–87.5%), and 86.0% (95% CI 81.8%–90.1%), respectively. Scoring 2 exhibited significantly superior performance than scoring 1 ( = 0.01) and scoring 3 ( < 0.001). Moreover, the accuracy of scoring 2 was not decreased significantly as compared to PI-RADS v2.1 ( = 0.05). There was no significant difference between 3 bpMRI-based scorings and with PI-RADS in TZ. However, although scoring 2 yielded the highest AUC, it was still notably inferior to PI-RADS ( = 0.02). All three bpMRI-based scorings demonstrated favorite diagnostic accuracy, and scoring 2 performed significantly better than the other two bpMRI-based scorings. Notably, scoring 2 was not significantly inferior to the full-sequence PI-RADS v2.1 in terms of sensitivity and specificity.