Medical researchers have developed a new prostate cancer prediction method by using machine learning tools to provide some of the highest levels of accuracy ever.
New Tools for an Old Problem
Prostate cancer has long been one of the deadliest forms of cancer afflicting men around the world. While treatment methods are getting better every year, early detection is still a major component of successful treatment; as a result, an emphasis on identifying prostate cancer while it’s still more easily treated has become a priority.
Using predictive modeling has always been part of early detection. Identifying risk factors means better screening options to look for early signs of prostate cancer in men that have those risk factors, and the quest for better prediction tools has led to some innovative research — including the results of a new study, recently published in the Journal of Scientific Reports, that details a screening method that makes use of machine learning to provide some of the most accurate predictive analysis results ever.
Pushing Detection and Classification Further
When it comes to assessing prostate cancer risk today, the primary tools are multiparametric MRI scans used in conjunction with a scoring system, called PI-RADS 2, for classifying anything detected in those scans. The problem with this standard method is that PI-RADS 2 tends to provide results that are highly subjective, making it difficult to tell whether a prostate tumor will be highly aggressive or just moderately so.
This uncertainty can lead to doctors choosing unnecessary treatment options for patients — or risking a patient’s health by misdiagnosing a tumor as less serious than it is. Knowing full well that more accurate detection and classification tools are needed to prevent such cases, researchers from USC’s Keck School of Medicine and Mount Sinai’s Icahn School of Medicine worked together to develop a tool that uses on machine learning.
This subset of artificial intelligence, which relies on algorithm-based data analysis, was found to predict the severity of tumors detected in multiparametric MRIs with better accuracy than traditional classification methods like PI-RADS 2 in a number of different cases. While further research will be needed to fine-tune this machine-learning approach, these early successes bode well for a future where predicting prostate cancer severity more accurately unlocks better, more appropriate treatment options that lead to higher treatment success rates for prostate cancer patients.
