If you are considering active surveillance (AS) for slow-growing prostate cancer, you really want to make sure that this is what you have: low-grade cancer cells with no aggressive features. The pathologist’s evaluation of your prostate biopsy tissue plays a critical role in determining your best course of action.

But not all pathologists are equal, says Hopkins pathologist Tamara Lotan, M.D. “The evaluation is often subjective, and there is significant variability among pathologists. Also, not much has changed in the way we have done this evaluation over the last 60 years.”

In a recent study, Lotan and her research team, along with Brady urologist Christian Pavlovich, M.D., and former Brady urologic oncology fellow Claire de la Calle, M.D., wondered whether artificial intelligence (AI) algorithms could improve the overall accuracy and consistency of prostate biopsy evaluations. The study was published in the Journal of the National Cancer Institute.

“We looked at how well these AI algorithms evaluated biopsy tissue of patients who are undergoing AS,” says Lotan. Their findings were impressive: “We found that in two separate AS cohorts,” men diagnosed with Grade Group 1 prostate cancer by the pathologist, “the AI algorithm sometimes assigned a higher grade to the tumor sampled in the initial diagnostic biopsy.” Then, the patients whose samples were “upgraded” by the AI algorithm “were more likely to be upgraded to Grade Group 2 or higher when their tumors were later re-biopsied than patients whose initial biopsy samples were not upgraded by AI.

“Thus, AI was better than the human eye at predicting which patients might subsequently require definitive therapy for their prostate tumors,” says Lotan. “Ultimately, such algorithms will give all AS patients access to expert pathology opinions, even if they do not get treated at large centers like Hopkins.”