On the heels of new recommendations for prostate cancer screening, researchers at Mount Sinai are studying how to more precisely diagnose the disease.
“In prostate cancer, the primary method that clinicians use to characterize a cancer and decide on a treatment protocol is the Gleason grade, or the Gleason score, of the patient,” says Gerardo Fernandez, MD, the Medical Director of Pathology at Mount Sinai St. Luke’s in New York. “Over the years, though, it turns out that we have over-treated prostate cancer by quite a bit. So the idea of having cancer leads a lot of patients to decide on radical treatments, namely surgery. What we’re trying to do is characterize prostate cancer more effectively than just by using the Gleason score.”
He adds: “Google might use similar techniques to characterize facial properties that allows them to do facial recognition software. We apply similar computer and engineering technology to images of cancer — how separated are the nuclei from one another, how bigger the nuclei, what protein markers do those nuclei contain. So we characterize these images with complex methods, and we extract these features out of them. Now, once we have the features, we use artificial intelligence-type approaches to take all of that data and figure out what combinations of features do a better job of predicting the behavior of the cancer than Gleason alone.”