Background MRI can noninvasively quantify blood flow by using phase-contrast imaging. Accurate flow measurements depend on placement of a “background” region of interest (ROI) on stationary tissue that is immediately adjacent to the flowing blood. This method does not work well in the heart and great vessels because there is little stationary tissue adjacent to the flowing blood. Therefore, flow measurements suffer from baseline offset errors due to artifacts from gradient eddy currents. The purpose of this study was to quantify the magnitude of the flow error and to propose a method to correct for this error.
Methods Blood flow in the ascending aorta and the main pulmonary artery was quantified in 10 healthy volunteers using MRI. Pulmonary blood flow (Qp) was measured in the main pulmonary artery just distal to the pulmonic valve. Aortic blood flow (Qs) was measured in the ascending aorta just distal to the coronary artery origins. (As such, aortic flow underestimates systemic blood flow by ≈5% because coronary flow is excluded. Consequently, for this study normal Qp/Qs is assumed to be 1.05.) To correct for baseline flow errors, immediately following the human subject's MR flow measurements, an identical MRI scan with the same parameters was performed on a stationary bottle of water (“phantom”). An ROI was placed on the phantom in the same image location where blood flow had been measured. Since there is no flow in a stationary phantom, nonzero flow values obtained from the phantom were assumed to be due to a baseline offset error, and this amount was subtracted from the flow of the corresponding ROI in the aorta or main pulmonary artery to correct the baseline flow error.
Results Before baseline correction, the measured Qp/Qs was 1.3 ± 0.2 (mean ± 1 SD, p < .01). After correction the measured Qp/Qs was 1.05 ± 0.07 (p = .89). Uncorrected difference between pulmonary and aortic flow was 26 ± 21 mL. After baseline correction with the phantom, the mean difference in flow was 7 ± 7 mL. Review of the measurements from all subjects showed a moderate correlation between the magnitude of error and the distance of the vessel from the center of the scanner (r2 = .55 for the aorta and r 2 = .61 for the pulmonary artery for offsets in the anterior-posterior direction).
Conclusion Baseline offset errors significantly affect MRI flow measurements in the heart and great vessels. These errors are large and are especially problematic for larger patients whose aortas lie further away from the center of the scanner. The correction method we propose substantially reduces the baseline offset. It thereby enables more accurate detection and quantification of blood flow, shunts, and valvular disease.