Detection of ISUP≥2 Prostate Cancers Using Multiparametric MRI: Prospective Multicenter Comparison of the PI-RADS Score and an Artificial Intelligence System

Multiparametric MRI of the prostate is recommended before each prostate biopsy. It identifies suspicious areas which will then be the subject of targeted biopsies. However, MRI suffers from low specificity and moderate inter-reader reproducibility, including with the use of the PI-RADS version 2.1 score. We are developing, within the framework of RHU PERFUSE, a computer-aided diagnosis system (CAD) for the detection of ISUP ≥2 cancers. This system has been trained on a database of patients who had prostate MRI and prostatectomy at the Hospices Civils de Lyon and performed well on a database of patients who had prostate MRI before biopsy at the Hopices civils de Lyon. However, one of the weaknesses of artificial intelligence systems is their low robustness when tested on MRI images from different manufacturers or institutions. The goal of the CHANGE study is to build a prospective multicenter cohort of patients who underwent prostate multiparametric MRI followed by systematic and targeted prostate biopsies. The cohort will be used for the final external validation of the CAD developed in PERFUSE.