Precision and accuracy of cross-sectional area measurements used to measure coronary endothelial function with spiral MRI.

Abstract

Coronary endothelial function (CEF) reflects vascular health and conventional invasive CEF measures predict cardiovascular events. MRI can now noninvasively measure CEF by quantifying coronary artery cross-sectional area changes in response to isometric handgrip exercise, an endothelial-dependent stressor. Area changes (10 to 20% in healthy; 2 to -12% in impaired vessels) are only a few imaging voxels because of MRI's limited spatial resolution. Here, with numerical simulations and phantom studies, we test whether Fourier interpolation enables sub-pixel area measurement precision and determine the smallest detectable area change using spiral MRI. In vivo coronary SNR with the currently used CEF protocol at 3T was measured in 7 subjects for subsequent in vitro work. Area measurements of circular vessels were simulated by varying partial volume, vessel diameter, voxel size, SNR, and Fourier interpolation factor. A phantom with precision-drilled holes (diameters 3-3.42 mm) was imaged 10 times with the current CEF protocol (voxel size, Δx = 0.89 mm) and a high-resolution protocol (Δx = 0.6 mm) to determine precision, accuracy, and the smallest detectable area changes. In vivo coronary SNR ranged from 30-76. Eight-fold Fourier interpolation improved area measurement precision by a factor 6.5 and 4.9 in the simulations and phantom scans, respectively. The current CEF protocol can detect mean area changes of 4-5% for SNR above 30, and 3-3.5% for SNR above 40 with a higher-resolution protocol. Current CEF spiral MRI with in vivo SNR allows detection of a 4-5% area change and Fourier interpolation improves precision several-fold to sub-voxel dimensions.