Researchers from Mount Sinai’s Icahn School of Medicine and Keck School of Medicine at the University of Southern California have developed a novel machine-learning framework that can precisely distinguish between low- and high-risk prostate cancer. The framework is intended to enable physicians, especially radiologists, to pinpoint treatment options for prostate cancer patients, decreasing potentially unnecessary clinical intervention.
Prostate cancer is a leading cause of cancer death in American men, second only to lung cancer. Although advances in prostate cancer research have saved many lives, objective prediction tools have remained an unmet need.
The current methods for assessing prostate cancer risk are multiparametric magnetic resonance imaging (mpMRI) and the Prostate Imaging Reporting and Data System, version 2 (PI-RADS v2), which classifies lesions found on the mpMRI. But PI-RADS v2 scoring is subjective and doesn’t clearly distinguish between intermediate and malignant cancer levels, leading to contrasting interpretations by clinicians.
Combining machine learning with radiomics has previously been suggested, but other studies only tested a small number of methods. However, by developing a predictive framework to rigorously and systematically assess many such methods, Mount Sinai and USC researchers were able to identify the best-performing one. They were also able to classify patients’ prostate cancer with high sensitivity and an even higher predictive value.
EurekAlert!, February 7, 2019