PREDICTIVE VALUE OF RADIOLOGIST EXPERIENCE IN INTERPRETING MULTIPARAMETRIC PROSTATE MRI DATA ACCORDING TO PI-RADS 2.1:
A RETROSPECTIVE SINGLE-CENTER STUDY
Rzaev R.T., He Mingze, Feldsherov M.V., Tychinskii A.S., Tyan A.S., Shchegelskii V.V.,
Zakharov R.S., Fedoseeva V.K.
Sechenov University. Moscow, Russia.
Purpose. To explore how radiologist experience influences the interpretation of prostate multiparametric MRI (mpMRI) with PI-RADS score 3 and its clinical utility in prostate cancer (PCa) diagnosis and management.
Materials and Methods. A retrospective single-center study included six radiologists with varying expertise levels to examine mpMRI images of patients designated with PI-RADS 3 scores. Out of 276 reviewed patients from January to December 2023, 35 met the inclusion criteria. Information on the localization of the pathological changes and the PI-RADS scores are reported according to the PI-RADS version 2.1 guideline. The results obtained were compared to analyze the inter-reader variability utilizing Cohen’s kappa (k) statistics. Kendall’s tau coefficient (t) was used for correlations between PI-RADS and Gleason scores, alongside various diagnostic performance metrics. Diagnostic performance metrics such as AUC, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated.
Results. Interpreting PI-RADS version 2.1, the median k score was 0.087, showing poor-to-fair agreement among radiologists. A weak positive correlation was found between PI-RADS scores and Gleason scores (t = 0.154; P = 0.549). Diagnostic performance varied, with Reader 1 showing the highest AUC (0.785) for superior PCa detection. Readers 2 and 4 had the highest accuracy (60%), while Reader 5 achieved 100% sensitivity. Readers 2 and 4 also recorded the highest specificity (91.9%).
Discussion. Six radiologists with varying levels of experience evaluated prostate multiparametric MRI (mpMRI) in accordance with the PI-RADSv2.1 guidelines. To assess inter-reader agreement, all specialists were blinded to prior mpMRI reports, histopathological results, and patient data. The expert radiologist achieved the highest AUC and PPV values in differentiating benign and malignant lesions; however, this reader did not demonstrate the highest or lowest sensitivity, specificity, accuracy, or NPV. The least experienced radiologist showed the poorest performance in terms of AUC, accuracy, NPV, and PPV among all readers, despite demonstrating the best correlation between imaging findings and pathological data. The obtained data suggest that the level of experience does not have a significant impact on diagnostic performance, highlighting the need for continuous training of less experienced specialists to improve the accuracy of prostate cancer detection. The observed inter-reader agreement was also low.
The PI-RADS 3 category remains a “gray zone” in clinical practice for both radiologists and urologists due to uncertainty regarding the presence of prostate cancer. In this study, eight patients were diagnosed with prostate cancer based on histological findings, accounting for 23% of the total sample. However, most of these cases were low-risk cancers with a Gleason score ≤ 7a (n = 7, 87.5%).
Large multicenter studies are needed to further validate the clinical value of PI-RADS 3 assessment.
Conclusion. The findings demonstrate poor-to-fair agreement among radiologists of differing experience levels when interpreting PI-RADS 3 scores using version 2.1 guidelines, suggesting that radiologist expertise does not significantly impact PI-RADS 3 score assessments. This underlines the need for larger, multicenter studies to enhance diagnostic consistency and potentially update guidelines for more precise PCa detection.
Keywords: PI-RADS version 2.1, Prostate cancer, multiparametric MRI, Inter-reader variability.
Corresponding author: Rzayev Ramin, e-mail: Этот e-mail адрес защищен от спам-ботов, для его просмотра у Вас должен быть включен Javascript
For citation: Rzaev R.T., He Mingze, Feldsherov M.V., Tychinskii A.S., Tyan A.S., Shchegelskii V.V., Zakharov R.S., Fedoseeva V.K. Predictive value of radiologist experience in interpreting multiparametric prostate MRI data according to PI-RADS 2.1: a retrospective single-center study. REJR 2026; 16(1):138-149. DOI: 10.21569/2222-7415-2026-16-1-138-149.
Received: 13.02.26 Accepted: 14.04.26