Video

Baris Turkbey, MD, discusses the transformative potential and challenges of artificial intelligence (AI) in prostate imaging and biopsies. He begins this 19-minute presentation detailing the benefits of MRI-guided prostate biopsies for localized cancer diagnosis. However, he highlights significant limitations, including low interobserver agreement among radiologists, undermining MRI’s reliability in broader clinical applications.

Turkbey explains how AI can address challenges by standardizing imaging quality and assisting in lesion detection. He references a large-scale European study (PI-CAI) demonstrating AI’s ability to detect clinically significant cancers. Importantly, he illustrates the critical yet underexplored interaction between AI and radiologists, showing that while AI often detects lesions missed by radiologists, its findings are sometimes disregarded, resulting in missed cancers.

Dr. Turkbey introduces an AI model developed at NCI that demonstrates high performance in lesion detection with low false positives. This model, employed for prospective biopsies, radiation oncology planning, and focal therapy, automates prostate segmentation and lesion identification processes, delivering results within minutes. However, his research reveals that human radiologists outperform AI in estimating tumor burden, particularly for PIRADS 4 lesions.