Using New Technology to Catch Prostate Cancer: Reading Biopsies with Artificial Intelligence
A recent study examined the accuracy of an artificial intelligence (AI) program to detect and
characterize prostate cancer in biopsy slides. The program from Ibex Medical Analytics that they used were more accurate than the programs examined in previous studies.
The researchers “trained” the AI algorithm to recognize prostate cancer with more than a million biopsy slide images. Expert pathologists had first labeled each image as healthy or abnormal tissue.
Then, they tested the AI algorithm on 1,600 slides from 100 patients with test results suggesting they may have prostate cancer. The AI demonstrated 98% sensitivity in detecting cancers and 97% specificity in distinguishing between cancer and no cancer.
The AI algorithm was the first to do more than just detect cancer. It also reported tumor grading, sizing, and invasion of the nerves within the prostate surrounding the cancerous regions.
The AI also flagged six slides with regions that were suspicious for cancer that the pathologists had not noted. While this could have been due to the pathologists seeing cancer elsewhere on the same patient’s slides, the AI could act as a failsafe to catch tricky cases such as these.
“Humans are good at recognizing anomalies, but they have their own biases or past experience,” said senior author Rajiv Dhir, M.D., M.B.A., chief pathologist and vice chair of pathology at University of Pittsburg Medical Center Shadyside and professor of biomedical informatics at University of Pittsburgh. “Algorithms like this are especially useful in lesions that are atypical. A nonspecialized person may not be able to make the correct assessment. That’s a major advantage of this kind of system.”