4D flow MRI facilitates detailed evaluation of cardiac hemodynamics in patients with cardiovascular disease. In this study, we investigated the performance of pseudo spiral compressed sensing (CS) accelerated whole-heart 4D flow MRI in a comparison with a clinically used EPI readout. CS-accelerated 4D flow MRI yielded similar results to EPI-accelerated 4D flow MRI in terms of velocity vector fields during ventricular ejection and filling and led to consistent blood flow measurements across heart valves. Our data suggest that CS 4D flow MRI has the potential to be accelerated even further for quantitative whole-heart hemodynamic imaging.
9 healthy volunteers (aged 24 ± 4y, 3 female) underwent cardiac MRI including whole-heart CS (n=9) and EPI 4D flow MRI (n=4) at 3T. 4D flow data were acquired in 30 cardiac phases during free-breathing with retrospective ECG-gating. Acquired and reconstructed spatial resolution were 3.0x3.0x3.0 mm3 and 2.8x2.8x3.0mm3 and three-directional VENC was set to 150 cm/s. Scan times ranged from 7 to 10 minutes, depending on FH-dimension of the (transversal oblique) FOV and the volunteer’s heart rate during the EPI scan. EPI was obtained with an EPI factor of 5 and a SENSE factor of 2. The CS undersampling factor ranged from 5.5 to 8.2 to keep the scan times for both scans the same. CS 4D flow scans were reconstructed offline with ReconFrame (Gyrotools, Zurich, Switzerland) and using the Berkeley Advanced Reconstruction Toolbox (BART) (5). A sparsifying total variation transform in time was used with regularization parameters of r = 0.001 and 20 iteration steps.
Time-averaged phase contrast MRA images were created from the EPI and CS datasets by multiplication of the magnitude with the absolute velocity images, and used to segment the entire heart including connected large arteries and veins (Mimics, Materialize, Leuven, Belgium). The overlap of both segmentations was maximized using rigid registration. CS velocity vectors were interpolated to the EPI segmentation for ventricular ejection and filling phases, followed by whole-heart voxelwise comparison for velocity magnitude and vector angle.
Semi-automated retrospective tracking of all four heart valves was performed using 2D cine bSSFP by dedicated software (CAAS MR 4D Flow 2v0, Pie Medical Imaging) with through-plane valve motion correction and automatic aliasing and phase offset correction.
Bland-Altman analysis was used to determine the mean difference and 95% limits of agreement (LOAs) between the CS and EPI velocity vectors. Furthermore, orthogonal regression was performed (Pearson correlation coefficient denoted by r).
Figure 1 shows an example of voxelwise comparison between CS and EPI velocity vector fields in a single volunteer at peak-ejection. Averaged over the cohort, a Pearson correlation of 0.87 ± 0.02 m/s was found at peak-ejection and 0.73 ± 0.06 m/s at peak-filling (Table 1). Bland-Altman analysis revealed a mean difference of -0.014 ± 0.028 m/s at peak-ejection and -0.001 ± 0.010 m/s at peak-filling, indicating no significant bias. Limits of agreement were in the order of 1/5 (peak-ejection) and 1/7 (peak-filling) of the VENC. Median angle difference was 16 ± 1 degrees for peak-ejection and 25 ± 2 degrees for peak-filling.
An example of an EPI and CS streamline visualization obtained with semi-automated retrospective valve tracking is shown in Figure 2. In Figure 3, measured flow curves in a single subject are displayed for both EPI and CS. The measurements revealed good consistency between heart valves (Figure 4).
1. Lustig M, et al., Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn. Reson. Med. 2007; 58:1182-1195
2. Gottwald LM, Peper ES, Zhang Q et al. Pseudo Spiral Compressed Sensing for Aortic 4D Flow MRI: a Comparison with k-t Principal Component Analysis. Proc 27rd Annu Meet ISMRM, Paris, France. 2018
3. Peper ES , Gottwald LM, Zhang Q et al. 30 times accelerated 4D flow MRI in the carotids using a Pseudo Spiral Cartesian acquisition and a Total Variation constrained Compressed Sensing reconstruction. Proc 27rd Annu Meet ISMRM, Paris, France. 2018
4. Westenberg JJ, Roes SD, Ajmone Marsan N, et al. Mitral valve and tricuspid valve blood flow: accurate quantification with 3D velocity-encoded MR imaging with retrospective valve tracking. Radiology. 2008;249:792–800.
5. Uecker M, et al. Berkeley advanced reconstruction toolbox. Proc. Intl. Soc. Mag. Reson. Med. 23; 2015