Artificial Intelligence Aids in Predicting Prostate Cancer
Two recent studies showcase how technology is helping NCI researchers better identify and understand prostate cancer.
- In one study, NCI researchers applied an artificial intelligence (AI) algorithm, which they developed in-house, to magnetic resonance imaging (MRI) scans to help identify cancerous lesions. Their results were on par with those of an expert radiologist and could be a critical step in validating the
algorithm for future clinical use. According to the authors, this tool may someday be an important patient triage tool—further assisting radiologists in reading diagnostic images for prostate cancer.
For an update on this model, see “A Prostate Imaging-Reporting and Data System Version 2.1-Based Predictive Model for Clinically Significant Prostate Cancer Diagnosis,” in BJU International.
- In a second study, a team of NCI researchers led by Dr. Stephanie A. Harmon developed an AI model to capture subtle details in tumor growth patterns from digital pathology images (from biopsies and prostate gland specimens). Their tool was better in predicting recurrence and survival (i.e., metastases-free survival), compared with the Gleason Scoring method that pathologists use today.
See the full article in the Journal of Pathology Informatics.
Dr. Baris Turkbey, an author on both studies, credits the technology underlying these important AI studies and their discoveries, noting, “These tools are excellent examples of how technology can help oncologists make more informed decisions to improve how we diagnose and care for people with prostate cancer. For example, using tools such as these we can determine when it’s appropriate to act aggressively in treating prostate cancer or if treatment warrants a more conservative approach.”
[callout]Want to learn more from Dr. Turkbey? Check out this episode of the Cancer HealthCast podcast where he discusses AI’s role in MRI-guided biopsies and identifies the AI tools that are helping researchers detect prostate cancer.