Coronary computed tomographic angiography (CCTA) has emerged as an effective non-invasive tool for detecting coronary artery disease (CAD) and identifying atherosclerotic plaque characteristics (APCs). CCTA studies have helped researchers understand the progression of atherosclerotic disease and identify plaque features that have predictive value for future cardiac events.
The clinical application of this data has allowed for improved diagnoses and optimized medical and interventional management. Patients with diabetes are at a higher risk for CAD and major adverse cardiovascular events. The CREDENCE trial found that stenosis and the plaque features that make up stenotic lesions were strongly predictive of downstream ischemia.
A new study published in Diabetes Care has shown that artificial intelligence (AI) can assist in analyzing coronary computed tomography angiography (CCTA) scans to detect coronary artery disease. The study used AI algorithms to analyze CCTA scans and identify atherosclerotic plaque components (APCs) in the coronary arteries.
The study, which was performed at multiple sites using single- or dual-source CT scanners of ≥64-detector rows, followed the guidelines established by the Society of Cardiovascular Computed Tomography. Patients received nitroglycerin to improve image quality, and β-blockers were administered to those who required heart rate control. The study found that image quality was acceptable in 99% of patients.
The AI-guided approach used automated analysis of CCTA using validated convolutional neural network models, including Visual Geometry Group (VGG) 19 network, 3D U-Net, and VGG Network Variant. These models used deep learning, a process based on AI-generated patterns of recognition and adaptation that are entirely AI-derived. Neural networks were used for lumen wall evaluation, vessel contour determination, and plaque characterization, optimizing for phase with each vessel segment analysis.
The software produced a centerline along the vessel’s length for lumen and outer vessel wall contouring. Vessels were then segmented and labeled by their position in the coronary tree and within the proximal, middle, or distal portion of the vessel itself. The software identified areas where the plaque was present by comparing a standard proximal cross-sectional reference slide to a regular distal cross-sectional reference slide on either end of a lesion.
These regular cross-sectional slides were then marked to signify a lesion’s proximal and distal ends. The software could then calculate the lesion’s length and the total plaque burden between the markers and characterize the present plaque components. Maximum stenosis was calculated by identifying the ratio of a regular cross-sectional slide with the slide that demonstrated the most remarkable luminal narrowing. The obstructive disease was defined by a lesion with ≥50% luminal narrowing compared with a regular cross-sectional reference slide.
Overall, the study suggests that AI-guided analysis of CCTA scans can provide a reliable and efficient way to detect coronary artery disease and identify patients who may require further evaluation or treatment. The approach may be beneficial in settings where ICA is unavailable or not feasible. The research, carried out across several medical centers, used CCTA (coronary computed tomography angiography) to measure arterial plaque composition in patients with and without diabetes.
The study found that patients with diabetes have a more significant burden of disease, higher PV, more diseased vessels, and higher NCP volume than those without diabetes. Furthermore, patients with diabetes and non-obstructive stenosis exhibited APCs (arterial plaque composition) similar to those with obstructive stenosis who did not have diabetes. The study has highlighted the importance of incorporating early plaque burden into ischemia prediction for patients with diabetes.
