Background: Two-dimensional (2D) perfusion cardiovascular magnetic resonance (CMR) remains limited by a lack of complete myocardial coverage. Three-dimensional (3D) perfusion CMR addresses this limitation and has recently been shown to be clinically feasible. However, the feasibility and potential clinical utility of quantitative 3D perfusion measurements, as already shown with 2D-perfusion CMR and positron emission tomography, has yet to be evaluated. The influence of systolic or diastolic acquisition on myocardial blood flow (MBF) estimates, diagnostic accuracy and image quality is also unknown for 3D-perfusion CMR. The purpose of this study was to establish the feasibility of quantitative 3D-perfusion CMR for the detection of coronary artery disease (CAD) and to compare systolic and diastolic estimates of MBF.
Methods: Thirty-five patients underwent 3D-perfusion CMR with data acquired at both end-systole and mid-diastole. MBF and myocardial perfusion reserve (MPR) were estimated on a per patient and per territory basis by Fermi-constrained deconvolution. Significant CAD was defined as stenosis ≥70% on quantitative coronary angiography.
Results: Twenty patients had significant CAD (involving 38 out of 105 territories). Stress MBF and MPR had a high diagnostic accuracy for the detection of CAD in both systole (area under curve [AUC]: 0.95 and 0.92, respectively) and diastole (AUC: 0.95 and 0.94). There were no significant differences in the AUCs between systole and diastole (p values >0.05). At stress, diastolic MBF estimates were significantly greater than systolic estimates (no CAD: 3.21 ± 0.50 vs. 2.75 ± 0.42 ml/g/min, p < 0.0001; CAD: 2.13 ± 0.45 vs. 1.98 ± 0.41 ml/g/min, p < 0.0001); but at rest, there were no significant differences (p values >0.05). Image quality was higher in systole than diastole (median score 3 vs. 2, p = 0.002).
Conclusions: Quantitative 3D-perfusion CMR is feasible. Estimates of MBF are significantly different for systole and diastole at stress but diagnostic accuracy to detect CAD is high for both cardiac phases. Better image quality suggests that systolic data acquisition may be preferable.