Cardiovascular disease remains the leading cause of death worldwide, and stable coronary artery disease (SCAD) accounts for the greatest proportion of cardiovascular disease. In the past decades, percutaneous coronary intervention (PCI) has become one of the most common treatments for SCAD, and therefore assessing the hemodynamic significance of coronary stenosis is important for physicians to make the optimal treating strategy. Coronary CT angiography (CTA) or invasive coronary angiography (CAG) is usually performed to evaluate the severity of coronary stenosis depending on the probability of CAD. However, the stenosis severity is not closely corresponding with the hemodynamic significance in coronary arteries.

As a result, fractional flow reserve (FFR) with pressure wire measurement was introduced to functionally assess the coronary stenosis. FFR is defined as the ratio of maximum blood flow distal to a stenotic lesion under hyperemia state to normal maximum flow in the same vessel. The cutoff value of FFR to detect significant ischemia is set to be 0.80, indicating that PCI should be considered if FFR0.80. FAME (Fractional Flow Reserve versus Angiography for Multivessel Evaluation) study confirmed that FFR guided PCI was superior to angiography guided PCI in reducing major adverse cardiovascular events (MACE) in patients with multivessel disease. In the subsequent FAME 2 study, FFR guided PCI plus the optimal medical treatment (OMT), as compared with the OMT alone, decreased the composite event rates mainly driven by urgent revascularization in SCAD patients. However, FFR does have some limitations, such as risks of pressure wire injury, extra time and cost, and side effects of hyperemic agents.

To overcome the limitations of FFR, CTA- and CAG-based methods to functionally assess coronary stenosis were proposed, i.e. FFR derived from CTA (FFRCT) and FFR derived from angiography-based quantitative flow ratio (QFR), which can simultaneously evaluate anatomic and hemodynamic significance of stenotic lesions. A number of studies have demonstrated that FFRCT has high sensitivity and specificity in identifying myocardial ischemia. However, the diagnostic accuracy of FFRCT depends on the image quality of coronary CTA, and it is relatively low in lesions with severe calcification and/or tortuosity. Besides, the methodology of FFRCT relies on computational fluid dynamics, which is complicated and time consuming. As for QFR, it is a novel method for deriving FFR based on 3-dimensional quantitative coronary angiography (3D-QCA) and contrast frame counting during CAG. Recent studies have shown that QFR has good diagnostic performance in evaluating the functional significance of coronary stenosis. The accuracy of QFR is also highly associated with anatomic information, thereby its diagnostic accuracy may be decreased in diffuse, tandem, thrombus-containing, calcified, or torturous lesions, and it is not suitable for prior infarction-related or collateral donor arteries as well. Given the above issues concerning FFRCT and QFR, we proposed a novel approach that integrates coronary CTA and CAG images to calculate FFR (FFRCT-angio) using artificial intelligence. The present study was undertaken to test the diagnostic accuracy of FFRCT-angio in patients with SCAD.