Artificial intelligence can exclude the presence of obstructive coronary artery disease (CAD) in patients who have undergone coronary artery calcium scoring, new research finds.
Coronary artery calcium scoring (CACS) is a reliable, low-dose method for estimating a patient’s buildup of plaque in the artery’s walls, and is strongly associated with all-cause mortality. The researchers found that combining deep learning with such measurements—along with additional cardiac analyses—was highly accurate at eliminating the presence of the disease, which can lead to a heart attack.
The team, from Silesian Center for Heart Diseases and Silesia Medical University in Poland, also believe their algorithm could increase the efficiency of ordering coronary computed tomography angiography (CCTA) exams for suspected CAD patients.
“Application of the proposed method can significantly decrease the number of patients referred to CTTA (by about 70%), which leads to the limitation of patients’ exposure to side effects of such tests and can reduce the costs of diagnostics for medical care providers,” Jan Glowacki, MD, PhD, with Silesian’s Heart Disease Center, and colleagues wrote.
Glowacki and co-researchers used data from more than 435 patient to create their algorithm. Each had undergone a clinically indicated CACS test between September 2017 and July 2018, and had a low to moderate probability of CAD. They then tested their model on data from 126 participants who received a CACS test from September to October 2018.
Overall, the algorithm has “considerably high” discriminatory power for ruling out obstructive CAD, the authors wrote. When tested on the 126 patient group, it produced 73 true negatives, zero false negatives, 20 true positives and 33 false positives.
This, Glowacki et al. argued, allows clinicians to understand which patients should be sent on to receive CCTA exams. And in the current analysis, the algorithm would have prevented 73 unnecessary scans, sparing patients unneeded radiation and a risk of contrast-induced renal failure, the team concluded.
The full study was published last month in Academic Radiology.