Sunday, November 26 | 9:30 a.m.-9:40 a.m. | S1-SSGU01-3 | Room S404
This session features an update on research toward clinical translation of AI in detecting clinically significant prostate cancer on MRI.
Motivated by a lack of validation of algorithms on multicenter datasets, a team from University College London in the U.K. led by presenter Francesco Giganti, MD, PhD, developed an AI-based application and then tested it on several external validation sets to assess real-world generalizability. They used the PROSTATEx public dataset for training the model and then tested it on several external validation sets, including data from five sites and six scanner models by two vendors with different field strengths (1.5 tesla/3 tesla) and acquisition protocols. Of the 794 patients included in the study, 34% had clinically significant prostate cancer.
The software automatically outputs scores intended to identify Gleason score ≥ 3+4 clinically significant prostate cancer. For selecting patients for biopsy, the AI identified patients with clinically significant prostate cancer at a sensitivity of 94%, specificity of 57%, negative predictive value of 95%, and an area under the curve of 0.85 using multiparametric MRI data from the blinded external validation set, according to Giganti and colleagues.
The researchers concluded that “the proposed AI solution shows comparable performance to radiologists in major expert studies, on a large real-world, multicenter external validation dataset with different scanners, vendors, field strengths and imaging protocols.”
Learn how they achieved these results during this Sunday morning session.