Prostate Cancer Coverage from Every Angle
Advertisement
Advertisement

Artificial Intelligence and the Detection and Grading of Prostate Cancer

By: Kayci Reyer
Posted: Tuesday, February 4, 2020

According to a population-based, diagnostic study published in The Lancet Oncology, artificial intelligence (AI) may be used to identify and diagnose prostate cancer in biopsy samples with clinically acceptable accuracy. More than 6,500 slides from needle-core biopsies from nearly 1,000 participants primarily served as the training ground for the novel AI system.

“Our results show that it is possible to train an AI system to detect and grade prostate cancer on the same level as leading experts,” noted Martin Eklund, PhD, of Karolinska Institutet in Sweden, in a Karolinska press release. “This has the potential to significantly reduce the workload of uropathologists and allow them to focus on the most difficult cases.”

In addition to the previously mentioned slides, the investigators used 271 digitized slides of biopsies from 93 patients who were not part of that trial. The AI system used these slides to train its biopsy assessment; then it evaluated an independent data set of 1,631 biopsies taken from 246 STHLM3 participants and a validation data set of 330 biopsies taken from 73 other men to detect and grade malignant tissues. Finally, a team of 23 uropathologists manually evaluated the system’s Gleason grading performance on a set of 87 biopsies.

The AI system was found to be accurate overall in identifying benign (n = 910) versus malignant (n = 721) biopsy cores in the independent data set, achieving an area under the receiver operating characteristic (ROC) curve (AUC) of .997. It had a comparable performance while identifying benign (n = 108) versus malignant (n = 222) biopsy cores in the validation data set, for an AUC of .986. When comparing the cancer length prediction by the AI system and the assignment of the reporting pathologists, there was a .96 correlation for the independent data set and .87 for the validation data set. In addition, the AI scored within the range of expert pathologists when assigning Gleason grades.

Disclosure: For full disclosures of the study authors, please visit thelancet.com.



By continuing to browse this site you permit us and our partners to place identification cookies on your browser and agree to our use of cookies to identify you for marketing. Read our Privacy Policy to learn more.