Attention: Restrictions on use of AUA, AUAER, and UCF content in third party applications, including artificial intelligence technologies, such as large language models and generative AI.
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.

JU INSIGHT Artificial Intelligence Improves the Ability of Physicians to Identify Prostate Cancer Extent

By: Sakina Mohammed Mota, PhD*, Avenda Health, Inc, Culver City, California; Alan Priester, PhD*, Avenda Health, Inc, Culver City, California, David Geffen School of Medicine, University of California, Los Angeles; Joshua Shubert, MS, Avenda Health, Inc, Culver City, California; Jeremy Bong, BS, Avenda Health, Inc, Culver City, California; James Sayre, PhD, University of California, Los Angeles; Brittany Berry-Pusey, PhD, Avenda Health, Inc, Culver City, California; Wayne G. Brisbane, MD, David Geffen School of Medicine, University of California, Los Angeles; Shyam Natarajan, PhD, Avenda Health, Inc, Culver City, California, David Geffen School of Medicine, University of California, Los Angeles; *co-first authors | Posted on: 17 Jul 2024

Mota SM, Priester A, Shubert J, et al. Artificial intelligence improves the ability of physicians to identify prostate cancer extent. J Urol. 2024;212(1):52-62.doi:10.1097/JU.0000000000003960

Study Need and Importance

In targeted therapy of prostate cancer, clinicians aim to treat clinically significant disease while minimizing damage to healthy tissue. However, conventional MRI underestimates the extent of disease, confounding treatment decisions. Artificial intelligence (AI) can help overcome this limitation, leveraging multimodal data to map cancer risk in 3D. This approach can reveal disease invisible to MRI, potentially improving patient selection, treatment planning, and oncologic efficacy. A multireader multicase study was conducted to compare physicians’ delineations of tumor extent using AI vs standard-of-care (SOC; ie, clinical judgement) and to evaluate AI’s impact on treatment decision-making.

What We Found

Ten physicians (7 urologists and 3 radiologists) trained in urologic oncology each evaluated 50 cases using both SOC and AI methods (1000 total evaluations). All patients were diagnosed with intermediate-risk prostate cancer via MRI-targeted biopsy then received radical prostatectomy. Whole mount pathology slides derived from surgical specimens were registered to MRI and used as ground truth (Figure). AI-assisted contours had significantly greater balanced accuracy (84.7% vs 67.2%) and sensitivity (97.4% vs 38.2%) than SOC. AI also achieved a substantially higher negative margin rate than SOC (72.8% vs 1.6%) and caused urologists to alter treatment decisions for 28% of cases.

image

Figure. Exemplary case from reader study showing artificial intelligence (AI) output (A), standard-of-care (SOC; orange), and AI-assisted cancer contours (blue; B), with corresponding ground truth from surgical pathology (C), and reader performance across all cases demonstrated improvement over SOC in all readers when using AI (D). PCa indicates prostate cancer.

Limitations

Study data were derived from radical prostatectomy cases from a single institution. Ground truth was defined by a single expert pathologist, with some potential subjectivity in tumor boundary annotation and Gleason grading.

Interpretation for Patient Care

AI cancer estimation is a promising improvement upon current practice. The enhanced accuracy of AI-derived tumor delineation could improve therapy planning, targeting, and oncologic efficacy. AI may even influence biopsy strategy, with perilesion (penumbra) cores sampled from higher-risk regions to improve tumor boundary sampling.

advertisement

advertisement