Skip to main content
Researchers find AI can detect clinically significant prostate cancer missed by human readers on MRI scans
Posted August 25, 2023




A team of researchers from Emory, Georgia Tech, and Case Western University have teamed up on a study to use artificial intelligence (AI) to provide greater early diagnosis of prostate cancer.


Typically, prostate cancer detection is sought using magnetic resonance imaging (MRI) scans, which are read by radiologists. Yet, the study’s findings found that certain significant cancer lesions invisible to a human reader can be identified using AI.


These results could drastically improve early detection and better outcomes for those who may have evidence of prostate cancer, the second most common cancer among men in the U.S., with approximately 1 in 8 men being diagnosed in their lifetime.


“Missing indolent lesions on an MRI scan is perhaps acceptable, but the challenge is when clinically significant lesions are missed on an MRI scan. Our study is important because it demonstrates the use of radiomic and AI tools for identifying those lesions that likely need to be aggressively treated but might be missed by a human reader on a MRI scans, ” says Anant Madabhushi, the study’s senior author and the Robert W. Woodruff Professor of Biomedical Engineering in the Wallace H. Coulter Department of Biomedical Engineering at Emory University and Georgia Tech, a member of the Cancer Immunology research program at Winship Cancer Institute of Emory University and Research Career Scientist at the Atlanta Veterans Administration Medical Center. 


Madabhushi partnered with another biomedical engineering professor, experienced radiologists and Lin Li, who earned her doctorate at Case Western Reserve University to examine prostate MRI datasets from 164 men with prostate cancer at four health systems.


We identified features of clinically significant prostate cancer that were quantitatively different between lesions that are visible and invisible on MRI. These features were then used to train and validate a machine learning model,” says Rakesh Shiradkar, PhD, co-author of the study recently published in the European Journal of Radiation Open and assistant professor in the Wallace H. Coulter Department of Engineering at Emory University School of Medicine and Georgia Institute of Technology.


The machine learning model was able to better distinguish between visible and invisible lesions on an MRI and offer a stronger predictor for prostate cancer than one using visible lesions alone. This paves the way for non-invasive, imaging-based diagnosis in the future. The research team plans to further validate the machine learning tool.


“By understanding the reasons behind the invisibility of these lesions, we hope to improve their detection and monitoring, leading to better management and patient outcomes,” says Li.



Kelly Petty  
Wallace H. Coulter Department of Biomedical Engineering