Abstract: Background and purpose: With the development of multi-parametric MR imaging techniques for detection of prostate carcinoma, medical imaging has shown the promising value in diagnosis, prediction of aggressiveness and evaluation of responses to different treatments. However, the increase of scanning sequences and the different weight of sequences in different regions of prostate put additional burden on diagnosticians, and thus the Prostate Imaging Reporting and Data System version 2 (PI-RADS V2) is generated. For popularizing PI-RADS V2, we performed this study to investigate interobserver agreement and accuracy in diagnosing prostate lesions. Methods: We retrospectively analyzed 98 patients with 141 prostate carcinoma lesions confirmed by biopsy and/or surgery who met the conditions of PI-RADS V2 assessment. Two readers independently assigned a PI-RADS V2 assessment category to the lesions. The Cohen’s kappa statistic was used to quantify interobserver agreement. The area under the receiver operating characteristic (ROC) curve (AUC) was calculated to determine reader accuracy for the detection of clinically significant prostate cancer (Gleason score ≥7). Simultaneously, the cutoff value of all ROCs were calculated, which would be regarded as the optimal value to define the prostate lesions as benign or malignant foci. Results: When a PI-RADS V2 assessment category ≥3 was considered positive, the agreement between readers was good for nonperipheral zone lesions (kappa=0.668) and peripheral zone lesions (kappa=0.769). When a PI-RADS V2 assessment category ≥4 was considered positive, the agreement was better for non-peripheral zone lesions (kappa=0.710) and excellent for peripheral zone lesions (kappa=0.843). The AUCs for readers 1 and 2 were 0.816 and 0.792, and had no significant difference. The AUCs were greater for non-peripheral zone lesions than for peripheral zone lesions (AUC1: 0.886 vs 0.791; AUC2: 0.791 vs 0.730). Additionally, the cutoff value of all ROC curves was 3. Conclusion: Two experienced readers were able to accurately identify patients with clinically significant prostate cancer using PI-RADS V2 with good interobserver agreement. PI-RADS V2 may be more suitable for non-peripheral zone lesions, which need further investigation.

Key words: Prostate Imaging Reporting and Date System version 2, Prostate carcinoma, Accuracy, Interobserver agreement