Phase contrast (PC) MRI is a useful tool for assessment of hemodynamic consequences caused by a variety of heart diseases. This study seeks to develop and test real-time PC MRI [1, 2] using compressed sensing (CS)[3, 4]. We chose to acquire breath-hold images in order to avoid any artifact due to motion. In this study, we compare two commonly used k-space sampling patterns (Radial and Cartesian). Both trajectories have desirable characteristics which may significantly affect PC reconstruction quality.Cartesian trajectories are less sensitive to gradient imperfection, and are capable of rectangular field of view that may be tailored for the human chest. Radial trajectories are more resistant to motion during acquisition [5], and have better SNR due to higher k-space center sampling. Furthermore, the temporal resolution of radial trajectories can be retrospectively determined when irrational angles such as golden angle ratio are used for acquisition [6]. To the best of our knowledge, no study has compared real time PC MRI between Cartesian and Radial k-space sampling schemes. We sought to compare their performance in human subjects.

(Imaging Acquisition) Based on the work by Joseph et al. [2], we implemented Radial real-time PC MRI pulse sequences using 7 k-space lines per image (i.e., 14 k-space lines per cardiac phase since interleaved reference and velocity-encoded acquisitions [7]). For real-time with Radial sampling, we used golden angle ratio with tiny gold angles [8](Fig. 1). For real-time with Cartesian sampling, we used a sampling pattern with variable density, where two central k-space lines are always sampled [8](Fig. 9). For reference, we performed breath-hold acquisition with GRAPPA (scan time = 9 heart beats). We acquired volunteers data on both 3T (Prisma, Siemens), and 1.5T (Espree, Siemens). The imaging parameters The three acquisitions (reference, radial real-time, Cartesian real-time) used the imaging parameters, which included: FOV between 350 – 400 mm for Cartesian acquisitions with 75% FOV in phase encoding direction, FOV between 275-300 mm for Radial acquisitions, slice thickness = 8 mm, flip angle = 20, real-time scan time = 6 heart beats, matrix size of 128X128 for Radial acquisitions, matrix size of 128X96 for Cartesian acquisitions, and temporal resolution = 74.70 ms for Espree data sets and 72.80 ms for Prisma data sets. We imaged a total of five valve planes. Two aortic valves, and two mitral valves on the Espreee, and 1 aortic valve on the Prisma. Aortic valves were images using venc = 250 cm/s and the mitral valve plane with venc = 150 cm/s. Pulse sequence order (reference, radial real-time, and Cartesian real-time) was randomized.

(Image Reconstruction) Coil sensitivity profiles were obtained by Self-Calibration [10], and Undersampled data sets were reconstructed using CS with temporal Total Variation (TTV) constraint in combination with SENSE parallel imaging (40 iterations). TTV normalized regularization weights were determined empirically using training date sets: Cartesian, 0.03; radial, 0.095 for Espree data sets, and Cartesian 0.04; radial, 0.6 for Prisma data set. Region of interests (ROI) were drawn manually around each valve, and the max velocity was calculated. Pearson’s correlation and Bland Altman analyses were conducted on pooled multi-phase velocity measurements

As shown in figure 2, compared with reference breath-hold acquisition, real-time PC MRI with Cartesian and Radial sampling schemes produced comparable image quality. Compared with radial, Cartesian sampling produced higher artifacts as shown. Compared with reference peak velocity measurements, both real-time PC MRI with Cartesian and radial sampling schemes produced measurements that are strongly correlated and in good agreement (see Fig. 3). This study demonstrates feasibility of real time PC MRI using either Cartesian and Radial k-space sampling pattern. Compared with Cartesian, radial produced less images artifacts. Future studies include assessment of diagnostic performance in patients with limited breath-hold capacity or in patients with arrhythmia (e.g., beat-to-beat variation).