Multiparametric Magnetic Resonance Imaging of the Prostate: Diagnostic Performance and Inter-reader Agreement of Two Scoring Systems

OBJECTIVE - To compare the diagnostic accuracies and inter-reader agreements of the Prostate Imaging Reporting and Data System 2 (PI-RADS v2) and University of California San Francisco (UCSF) multiparametric prostate MRI scale for diagnosing clinically significant prostate cancer.

METHODS - This IRB-approved retrospective study included 49 men who had 1.5-T endorectal MRI and prostatectomy. Two radiologists scored suspicious lesions on MRI using PI-RADS v2 and the UCSF scale. Percent agreement, 2x2 tables, and the area under the receiver operating characteristic curves (Az) were used to assess and compare the individual and overall scores of these scales. Inter-reader agreements were estimated with kappa statistics.

RESULTS - Reader 1 (R1) detected 78 lesions, reader 2 (R2), 80. Both identified 52 of 65 significant cancers. The Az for PI-RADS v2 and UCSF scale for R1 were 0.68 and 0.69 (T2WI), 0.75 and 0.68 (DWI), and 0.64 and 0.72 (overall score), respectively. And they were 0.72 and 0.75 (T2WI), 0.73 and 0.67 (DWI), and 0.66 and 0.75 (overall score) for R2. The DCE percent agreements between scales were 100% (R1) and 95% (R2). PI-RADS v2 DWI of R1 performed better than UCSF DWI (Az = 0.75, Az = 0.68; P = 0.05); no other differences were found. The inter-reader agreements were higher for PI-RADS v2 (T2WI: 0.56 vs. 0.42; DWI: 0.60 vs. 0.46; overall: 0.61 vs. 0.42). The UCSF approach to derive the overall PI-RADS v2 scores increased the Az for the identification of significant cancer (R1 to 0.76, P < 0.05; R2 to 0.71, P = 0.35).

CONCLUSIONS - While PI-RADS v2 DWI score may have a higher discriminatory performance than the UCSF scale counterpart to diagnose clinically significant cancer, the utilization of the UCSF scale weighing system for the integration of PI-RADS v2 individual parameter scores improved the accuracy its overall score. Advances in Knowledge: PI-RADS v2 is moderately accurate for the identification of clinically significant prostate cancer, but the utilization of alternative approaches to derive the overall PI-RADS v2 score, including the one used by the UCSF system, may improve its diagnostic accuracy.

The British journal of radiology. 2016 Mar 23 [Epub ahead of print]

Wei-Ching Lin, Valdair F Muglia, Gyl E B Silva, Salomão Chodraui Filho, Rodolfo B Reis, Antonio C Westphalen

1 Department of Radiology and Biomedical Imaging, University of California, San Francisco., 3 Department of Internal Medicine / Division of Radiology, Ribeirão Preto School of Medicine, University of São Paulo., 4 Department of Pathology, Ribeirão Preto School of Medicine, University of São Paulo., 3 Department of Internal Medicine / Division of Radiology, Ribeirão Preto School of Medicine, University of São Paulo., 5 Department of Surgery / Division of Urology, Ribeirão Preto School of Medicine, University of São Paulo., 6 Departments of Radiology and Biomedical Imaging, and Urology, University of California, San Francisco.