Establishing the Clinical Feasibility of a Transparent Computed-Tomography Based Fractional Flow Reserve (CT-FFR) Algorithm

Abstract

Coronary artery disease, a condition where plaque buildup in a coronary artery restricts blood flow to the myocardium, is a leading cause of heart disease, which is responsible for a high mortality rate in the US each year. While stress testing and echocardiography have traditionally been the gatekeepers for referrals to invasive coronary angiography (ICA) with possible revascularization, the inclusion of coronary computed tomography angiography (CTA) in risk assessment has been shown to reduce the number of normal coronary arteries found at ICA, i.e. false positives. Coronary CTA, however, provides only anatomic information. An important physiologic metric that has grown to become the gold standard in evaluation of lesion hemodynamic significance is fractional flow reserve (FFR). While FFR has traditionally been recorded in the cath lab, deriving FFR from coronary CTA imaging (CT-FFR) is gaining traction as a desirable noninvasive alternative. Already, CT FFR has been shown in various clinical trials to have superior diagnostic performance compared to CT anatomic analysis of stenosis alone. There are four essential components to the CT-FFR computational workflow: 1) CTA image segmentation and 3-D coronary tree reconstruction 2) determination of boundary conditions, including estimation of total coronary flow at rest 3) simulation of hyperemic conditions, including accounting for changes in microvascular and/or epicardial coronary resistance, and distribution of blood flow among various branches and finally 4) solving Navier-Stokes fluid dynamics equations with a computational fluid dynamics (CFD) solver. The purpose of this study was to examine how variations in several of these components individually affected the diagnostic accuracy of an open, non-proprietary CT-FFR algorithm developed and validated by the Applied Imaging Science Laboratory (AISL) at Brigham and Women’s Hospital that can be performed on a standard radiology workstation in 1 hour, using invasive FFR measurements as the reference standard. A secondary goal of this study was to investigate whether endothelial shear stress, an alternative metric that does not require hyperemic simulation, is associated with FFR and can therefore be used in lesion risk assessment. A retrospective study in 61 patients at a single medical center in Japan who underwent CTA followed by ICA show that the computational CT-FFR algorithm is fairly robust to changes in patient-specific and image specific characteristics, with a few exceptions. Optimal inlet flow boundary conditions for achieving the best CT-FFR diagnostic performance involve scaling total flow to myocardial mass. In addition, endothelial shear stress in the segment 5 mm distal to the plaque center is significantly correlated with FFR. High ESS in this segment appears to be a high-risk marker for hemodynamically significant FFR. These findings are useful for enhancing the accuracy of future CT-FFR calculations because they shed light on the optimal set of parameters that should be inputted in the CT-FFR algorithm and further establish the clinical feasibility of CT-FFR as an efficacious gatekeeper to invasive testing, able to identify those with functionally obstructive coronary artery stenosis but also reduce the number of normal arteries referred to angiography.