Deep-Learning System Can Diagnose Prostate Cancer
Study examines performance of an automated deep-learning system in Gleason grading of prostate cancer using biopsies
Today’s Clinical Lab is a reader-centric publication that keeps clinical professionals up to date with today’s rapidly changing lab industry with in-depth and timely editorial content and resources, including clinical industry news and insights into the latest trends, technologies, and techniques in the clinical lab.
ViewFull ProfileLearn about ourEditorial Policies.
Researchers have developed a deep-learning system to grade aggressiveness of prostate cancer based on biopsies following the Gleason grading standard. According to a study published January 8, 2020 in The Lancet Oncology, the new system achieved a level of performance similar to that of pathologists, making it potentially useful in prostate cancer diagnosis.
To train the system, the researchers exposed it to hundreds of biopsy images already classified by expert urological pathologists. They then collected 5759 biopsies from 1243 patients at the Radboud University Medical Center in the Netherlands and compared the performance of the deep-learning system to that of a panel of 15 pathologists from different countries and with varying levels of experience.
The system outperformed 10 of the 15 pathologists, with its ability to grade biopsies comparable to highly experienced pathologists, the researchers report. They conclude that this deep-learning system could be implemented to assist pathologists by screening biopsies, providing second opinions on grade group, and measuring volume percentages.