NORFOLK, VIRGINIA - Prostate cancer is the second leading cause of cancer death in men, affecting about one in eight. That more men will die with prostate cancer than from prostate cancer is a frequent statement. Statistically, one in 41 men will die of prostate cancer, about one-fifth of those diagnosed. Couple these statistics with treatments that have a high rate of morbidity, and the need for high quality diagnostic and prognostic tools to differentiate the indolent tumors from cancers needing aggressive treatment becomes clear. Here are a few ways that AI advances are improving accuracy and efficiency in prostate imaging.

Prostate imaging’s most used modalities are magnetic resonance imaging (MRI), positron emission tomography (PET) and computerized tomography (CT). These imaging studies provided data used for prognostication, surgical planning, and radiation therapy mapping. Research groups have developed several AI algorithms to assist in the interpretation of these scans, mostly focused on improving accuracy and consistency among readers. The following is a brief discussion of current advances in the academic literature broken down into PET/CT and MRI, followed by a rundown of the US Food and Drug Administration (FDA) approved algorithms for prostate imaging in the commercial space.

PET/CT

PET is a common staple of oncologic imaging. This imaging modality utilizes positron emitting elements attached to tracers. Some of the many tracers include F-18 fluorodeoxyglucose (FDG) - a glucose analog that gives information on relative metabolic activity of tissues, F-18 sodium fluoride - whose biodistribution correlates with osteoblastic activity, and prostate specific membrane antigen (PSMA) labeled with Galium-68.

Commonly used metrics in PET image analysis are the Standardized Uptake Value (SUV) maximum (or max) and peak. SUV max and