| Summary: | Aim
The purpose of this study is to compare PI-RADSv2 with v2.1 in terms of interreader agreement and diagnostic performance in clinically significant prostate cancer (CSCa) detection.
Materials and Methods
Two hundred-two patients who had both multiparametric prostate magnetic resonance imaging (mpMRI) and 12 quadrant systematic biopsies were included in this retrospective study. Acquisition parameters were totally complied with proposal of PI-RADSv2 and 2.1 guidelines. mpMRIs were evaluated by two radiologists independently. Index lesion’s score was used to determine diagnostic performance of the systems. Gleason ≥ 3+4 tumors were considered CSCa. Kappa statistic was used to determine interreader agreement. Area under the curve (AUC) was calculated in detection of CSCa using receiver operating characteristic curve (ROC).
Results
The numbers of cases with PI-RADSv2 and v2.1 scores from 1 to 5 were 46, 21, 13, 41, 81 and 46, 26, 9, 43, 78, by reader 1 and 51, 10, 18, 43, 80 and 51, 15, 19, 37, 80 by reader 2, respectively. There was “substantial” agreement between two readers for both scoring systems. Kappa values were 0.600 in PI-RADSv2 and 0.624 in PI-RADSv2.1. Ninety-eight (48.5%) patients had CSCa. AUC values of PI-RADSv2 and v2.1 scoring systems were 0.861 and 0.851 for reader 1, 0.873 and 0.883 for reader 2, respectively.
Conclusion
Interreader agreement was “substantial” in mpMRI and slightly improved with PI-RADSv2.1. Diagnostic performance of the two systems were almost equal.
|
|---|