"Millions of chest CT scans are taken each year, often in healthy people, for example to screen for lung cancer. Our study shows that important information about cardiovascular risk is going unnoticed in these scans," said senior author Hugo Aerts, PhD, director of the Artificial Intelligence in Medicine (AIM) Program at Mass General Brigham. "Our study shows that AI has the potential to change how clinicians practice medicine and enable physicians to engage with patients earlier, before their heart disease advances to a cardiac event."

Chest CT scans can detect calcium deposits in the heart and arteries that increase the risk of a heart attack. The gold standard for quantifying CAC uses "gated" CT scans, that synchronize to the heartbeat to reduce motion during the scan. But most chest CT scans obtained for routine clinical purposes are "nongated."

The researchers recognized that CAC could still be detected on these nongated scans, which led them to develop AI-CAC, a deep learning algorithm to probe through the nongated scans and quantify CAC to help predict the risk of cardiovascular events. They trained the model on chest CT scans collected as part of the usual care of veterans across 98 VA medical centers and then tested AI-CAC's performance on 8,052 CT scans to simulate CAC screening in routine imaging tests.

The researchers found the AI-CAC model was 89.4% accurate at determining whether a scan contained CAC or not. For those with CAC present, the model was 87.3% accurate at determining whether the score was higher or lower than 100, indicating a moderate cardiovascular risk. AI-CAC was also predictive of 10-year all-cause mortality -- those with a CAC score of over 400 had a 3.49 times higher risk of death over a 10-year period than patients with a score of zero. Of the patients the model identified as having very high CAC scores (greater than 400), four cardiologists verified that almost all of them (99.2%) would benefit from lipid lowering therapy.

"At present, VA imaging systems contain millions of nongated chest CT scans that may have been taken for another purpose, around 50,000 gated studies. This presents an opportunity for AI-CAC to leverage routinely collected nongated scans for purposes of cardiovascular risk evaluation and to enhance care," said first author Raffi Hagopian, MD, a cardiologist and researcher in the Applied Innovations and Medical Informatics group at the VA Long Beach Healthcare System. "Using AI for tasks like CAC detection can help shift medicine from a reactive approach to the proactive prevention of disease, reducing long-term morbidity, mortality and healthcare costs."

Limitations to the study include the fact that the algorithm was developed on an exclusively veteran population. The team hopes to conduct future studies in the general population and test whether the tool can assess the impact of lipid-lowering medications on CAC scores.

Authorship: In addition to Aerts, Mass General Brigham authors include Simon Bernatz, and Leonard Nürnberg. Additional authors include Raffi Hagopian, Timothy Strebel, Gregory A. Myers, Erik Offerman, Eric Zuniga, Cy Y. Kim, Angie T. Ng, James A. Iwaz, Sunny P. Singh, Evan P. Carey, Michael J. Kim, R. Spencer Schaefer, Jeannie Yu, and Amilcare Gentili.

Funding: This work was funded by the Veterans Affairs health care system.