In recent years, studies have shown that coronary plaque burden is the main driver of cardiovascular outcomes and cardiac ischemia is likely a surrogate for coronary plaque burden.1,2 Histological studies have consistently shown high-risk or vulnerable plaque has features such as large volume, a lipid-rich necrotic core, positive remodeling, hemorrhage, peripheral neovascularization, a thin fibrous cap and chronic inflammation.3 Computed tomography (CT) evaluation of coronary plaque features can also identify lipid-rich necrotic core, which appears as low attenuation (<30 Hounsfield Units) plaque. Low-attenuation coronary plaque (LAP) volume in coronary Computed Tomography Angiography (CCTA) was found to be a strong predictor of non-fatal myocardial infarction and cardiovascular mortality at 5 years.4,5 More recently, identifying the presence of pericoronary adipose tissue (PCAT) with CT have been associated with coronary inflammation and improved risk stratification.6

To assess coronary arterial plaque burden and plaque characteristics on CT, several post-processing software programs have been developed and are currently commercially available.7,8 Some of these programs are cloud-based and use artificial intelligence (AI) to automatically segment the coronary vessels, label the vessel lumen and wall, quantify atherosclerotic plaque burden and plaque features, and assess degree of stenosis.9 There is limited data on intra- and inter-reader reproducibility in the quantification of calcified plaque, non-calcified plaque, and PCAT using a semi-automated AI-enhanced post-processing software. Some studies suggest high variability and discrepancy between artificial intelligence (AI)-enhanced approaches when compared to human expert readers.10,11 The objective of the current study was to determine the intra- and inter-reader reproducibility in quantifying calcified plaque, LAP and PCAT in 10 African American women both with or without human immunodeficiency virus (HIV), who were participating in the Specialized Center of Research Excellence (SCORE) clinical trial (5U54AG062334) and 6 men. This evaluation was conducted by two expert readers employing an AI-enhanced semi-automated CT analysis software.