Evaluation of the PI-RADS v2.1 Score Using Multiple Readers

The interpretation of prostate multiparametric MRI (mpMRI) is difficult and requires expertise. As a result, it suffers from substantial inter-reader variability. The so-called Prostate Imaging Reporting and Data System (PI-RADS) scoring system has been launched in 2012 to try and standardise prostate mpMRI interpretation. It is a 5-level score that assesses the likelihood that suspicious focal prostatic lesions seen on mpMRI are clinically significant prostate cancers. Despite the use of semi-objective criteria for each category of the score, the inter-reader reproducibility of the first two versions (PI-RADS v1 launched in 2012 and PI-RADS v2 launched in 2015) was moderate at best, even for experienced readers. The last version (PI-RADS v2.1) has been launched in March 2019 in an effort to improve the inter-reader reproducibility. This version has not been evaluated yet. The purpose of our study is to evaluate the accuracy and inter-reader reproducibility of the PI-RADS v2.1 score on a large set of 171 prostate MRIs using 21 readers of varying experience. Twenty-one readers (14 seniors and 7 juniors) from 9 different institutions and with varying experience in prostate mpMRI accepted to participate to the study. Reader will assess the dataset independently and will be blinded to the other readers' results. They also be blinded to clinical and biochemical data.