To evaluate the diagnostic performance and inter-reader reliability of the multiparametric magnetic resonance imaging (mpMRI) based prostate imaging reporting and data system (PI-RADS) version 1 and version 2 for the assessment of prostate cancer.A cohort of 82 patients underwent endorectal mpMRI at 1.5T. Patients had at least one lesion with a PI-RADS v1 assessment category of ≥3 and were selected for targeted in-bore MR-guided biopsy in a subsequent session. The results of the histopathological workup were used as reference standard. All lesions were retrospectively evaluated according to PI-RADS v2 by an experienced and unexperienced blinded reader. Diagnostic performance was compared by analyzing the area under the Receiver Operating Characteristics Curve (AUC). The weighted kappa method was used to calculate inter-reader reliability.Targeted MR-guided biopsy was performed in 136 lesions and revealed 39 malignant lesions in 31 patients. AUC values increased for the experienced reader (PI-RADS v1 0.79; PI-RADS v2 0.83) and unexperienced reader (PI-RADS v1 0.70; PI-RADS v2 0.83). When excluding the cases of low grade cancer (Gleason score=3+3), AUC values increased further for the experienced reader (PI-RADS v1 0.88; PI-RADS v2 0.91) and unexperienced reader (PI-RADS v1 0.78; PI-RADS v2 0.90). Specificity at the selected threshold of a PI-RADS v1/v2 assessment category ≥4 improved for both readers. Inter-reader agreement increased from κ=0.55 in PI-RADS v1 to κ=0.68 in v2.PI-RADS v2 improved diagnostic performance for the assessment of suspicious intraprostatic lesions identified in PI-RADS v1 for both readers and led to higher inter-reader reliability. These results suggest that PI-RADS v2 is a reliable and replicable reporting system for the assessment of prostate cancer.