Deep learning validated on 20K patients uses CT scans to predict cardiovascular risk
Researchers have developed a new deep learning system that measures coronary artery calcium from CT scans to help providers estimate an individual’s risk of experiencing an adverse cardiovascular event, such as a heart attack.
Artificial intelligence researchers at Brigham and Women’s Hospital teamed up with Massachusetts General Hospital cardio imaging experts to develop their system. And after putting it to the test, they found its auto-generated calcium scores matched those calculated manually by a human expert.
The tool was validated on more than 20,000 individuals and may help physicians and patients to make more informed decisions about caring for their hearts.
"Coronary artery calcium information could be available for almost every patient who gets a chest CT scan, but it isn't quantified simply because it takes too much time to do this for every patient," Hugo Aerts, PhD, director of the Artificial Intelligence in Medicine Program at the Brigham and Harvard Medical School, said in a statement. "We've developed an algorithm that can identify high-risk individuals in an automated manner."
Aerts and colleagues trained and tested the deep learning system on data from a handful of National Heart, Lung and Blood Institute-funded imaging and outcome trials. These included the Framingham Heart Study, Prospective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE), and the Rule Out Myocardial Infarction using Computer Assisted Tomography trial (ROMICAT-II). Data from the National Lung Screening Trial (NLST) was also used.
In addition to matching manually developed CAC scores, the automated scores also independently predicted who would have a major adverse cardiovascular event, such as a heart attack.
Currently, this tool is only ready for research purposes but is open source for others to test. But the team believes it has a very bright future ahead.
“This is an opportunity for us to get additional value from these chest CTs using AI," added co-author Michael Lu, MD, MPH, director of artificial intelligence at MGH's Cardiovascular Imaging Research Center. “From a clinical perspective, our long-term goal is to implement this deep learning system in electronic health records, to automatically identify the patients at high risk."
Read the entire study published Jan. 29 in Nature Communications here.