Patients getting chest CT scans for lung cancer screenings can also be measured for heart disease thanks to AI, according to a new study presented at RSNA 2019 in Chicago.
That’s because coronary artery calcium, a measurement of plaque in arteries, can be visible on low-dose CT scans. These scans are generally intended to diagnose lung cancer among high-risk patients, but the coronary artery calcium scored derived from CT can help doctors prescribe statins, or cholesterol-lowering preventive medications.
For example, patients with a calcium score of 0 can defer statin treatment, but patients with a high calcium score should be on a statin, according to study co-senior author Michael T. Lu, MD, MPH, director of AI in the Cardiovascular Imaging Research Center at Massachusetts General Hospital (MGH) in Boston.
"The new cholesterol guidelines encourage using the calcium score to help physicians and patients decide whether to take a statin," Lu said in a prepared statement.
Boston researchers from MGH and the AI in Medicine program at Brigham and Women’s Hospital developed and tested a deep learning technique to measure coronary artery calcium on chest CT images, which is not routinely measured in these scans because it requires dedicated software and time to interpret. By comparison, the AI system developed by the researchers requires no extra exam time and runs in the background.
The AI tool, which was trained on cardiac and chest CTs where the coronary artery calcium was measured manually, determined calcium scores on par with human readers. The system was also tested on CT scans from thousands of heavy smokers between the ages of 55 and 74 from the National Lung Screening Trial.
Researchers also found deep learning calcium scores were associated with cardiovascular death over a 6.5-year follow-up.
"There's information about cardiovascular health on these CT scans," Lu said. "This is an automated way to extract that information, which can help patients and physicians make decisions about preventative therapy."
The tool could help examine large numbers of people for heart disease faster than human readers. It also has implications outside lung cancer screenings, showing effectiveness for people with stable and acute chest pain.