Expanding the coronary tree reconstruction to smaller arteries improves the accuracy of FFRCT

Objectives We attempted to improve the accuracy of coronary CT angiography (CCTA)-derived fractional flow reserve (FFR) (FFRCT) by expanding the coronary tree in the computational fluid dynamics (CFD) domain. An observational study was performed to evaluate the effects of extending the coronary tree analysis for FFRCT from a minimal diameter of 1.2 to 0.8 mm. Methods Patients who underwent CCTA and interventional FFR were enrolled retrospectively. Seventy-six patients qualified based on the inclusion criteria. The three-dimensional (3D) coronary artery tree was reconstructed to generate a finite element mesh for each subject with different lower limits of luminal diameter (1.2 mm and 0.8 mm). Outlet boundary conditions were defined according to Murray's law. The Newton-Krylov-Schwarz (NKS) method was applied to solve the governing equations of CFD to derive FFRCT. Results At the individual patient level, extending the minimal diameter of the coronary tree from 1.2 to 0.8 mm improved the sensitivity of FFRCT by 16.7% (p = 0.022). This led to the conversion of four false-negative cases into true-positive cases. The AUC value of the ROC curve increased from 0.74 to 0.83. Moreover, the NKS method can solve the computational problem of extending the coronary tree to an 0.8-mm luminal diameter in 10.5 min with 2160 processor cores. Conclusions Extending the reconstructed coronary tree to a smaller luminal diameter can considerably improve the sensitivity of FFRCT. The NKS method can achieve favorable computational times for future clinical applications.

Automated summary

By expanding the coronary tree in the computational fluid dynamics (CFD) domain, the accuracy of coronary CT angiography (CCTA)-derived fractional flow reserve (FFR) (FFRCT) was significantly improved, leading to a 16.7% increase in sensitivity and the conversion of false-negative cases into true-positive cases. Additionally, the NKS method was able to solve the computational problem efficiently, achieving favorable computational times for future clinical applications.