Upcoming Site Maintenance on Tuesday, May 28, 2024: Please note that some site functionality such as new user registrations, updates to user accounts, and article purchases will be unavailable during maintenance on this date.
You are prohibited from using or uploading content you accessed through this website into external applications, bots, software, or websites, including those using artificial intelligence technologies and infrastructure, including deep learning, machine learning and large language models and generative AI.


Utilization of multiparametric (mp) MRI and targeted biopsy is rapidly increasing. MRI level of suspicion (PIRADS score) is a powerful predictor of biopsy outcome. All published results have come from centers where 1-3 sub-specialized radiologists read all prostate MRIs. To enable widespread clinical utility, MRI interpretation must be reproducible across radiologists. At Stanford, 10 radiologists specialized in abdominal imaging read prostate MRI. We evaluated the consistency in MRI interpretation and biopsy outcome across radiologists.


We identified the first 206 men who underwent mpMRI and targeted biopsy in our prospective biopsy database. mpMRI included T2, diffusion, and contrast enhanced imaging acquired at 3-Tesla. MRI interpretation used PIRADS version 1. One urologist performed all biopsies. Age, PSA, biopsy indication, lesion PIRADS score, radiologist, and biopsy result were evaluated for each subject. The PI-RADS score distribution and biopsy cancer yield per lesion were analyzed for each radiologist.


Interpretation of 206 MRIs was unevenly divided among ten radiologists of varying experience (median = 27 MRI per radiologist, range 7-37). 243 lesions were identified overall (37% PIRADS 3, 41% PIRADS 4, 23% PIRADS 5). Cancer was present in 20% of PIRADS 3 lesions, 53% of PIRADS 4, and 95% of PIRADS 5. Substantial variability existed in cancer yield for PIRADS 3 and 4 lesions across radiologists. Cancer yield ranged from 0-50% for PIRADS 3 and from 20-100% for PIRADS 4 (figure). Similar variability existed for significant cancers (Gleason ≥7). Conversely, cancer yield for PIRADS 5 lesions was consistent across radiologists (range 80-100%). Odds ratios for correct identification of cancer differed significantly between radiologists despite controlling for age, PSA and biopsy indication.


We identified substantial variation in diagnostic accuracy across radiologists despite a constant MRI protocol, adherence to PIRADS, and biopsy performance by the same urologist. Targeted biopsy cancer yield depended both on PIRADS score and on the interpreting radiologist. Using many radiologists of varying experience may make our results more generalizable to the community. Variability across radiologists hampers clinical decision making. Greater efforts to standardize MRI interpretation are needed.

Download PPT