Comparison of kt-SENSE and kt-GRAPPA applied to Cardiac Cine and Phase Contrast Imaging

Introduction: To reduce scan time in time-resolved imaging, advanced parallel imaging techniques have been introduced such as kt-SENSE [1], kt-GRAPPA [2], and PEAK-GRAPPA [3] as an extension of ktGRAPPA. In this work, kt-SENSE was compared to PEAK-GRAPPA (below referred as kt-GRAPPA) in terms of error, noise behavior and temporal fidelity for cardiac Cine and phase contrast (PC) data. Methods: Measurements were performed on a 3T Trio system (Siemens) with a 12-channel thorax coil. In vivo time-resolved cardiac measurements in a healthy volunteer were performed using a bSSFP sequence (TR=3 ms) with 26 time frames (temporal res. 36 ms) and a matrix size of 192 x 256 acquired during breath-hold in a short axis orientation. Subsequently, an additional noise-only data set with zero flip angle was measured to allow for SNR analysis. The PC measurement was performed (TR=4 ms) with 16 cardiac time frames (temporal res. 64 ms) and a matrix size of 192 x 256 acquired during breath-hold in the aortic outflow tract with one-directional velocity encoding (through-plane venc=1.5 m/s). To compare the full data to kt-SENSE and kt-GRAPPA, k-space lines were removed retrospectively from the fully acquired data sets. Cardiac bSSFP and phase contrast data sets were reconstructed with a reduction factor of R=6 and 12 reference lines for both k-t-SENSE and kt-GRAPPA yielding a net acceleration factor of Rnet=4.6. Global and regional (encompassing LV for cardiac and aorta for PC) rootmean-square errors between the accelerated and full data averaged over all time frames were determined. The distribution of noise in the kt-SENSE and kt-GRAPPA images was determined by feeding noise samples into reconstruction while still using the reference lines from the Cine scan. Phased array image combination was performed as described by Roemer [4] with coil sensitivities calculated as temporal average derived from the fully acquired k-space data. For the cardiac data set SNR was estimated in a ROI within the myocardium (1 in Fig.2) and in static muscle tissue (2 in Fig.2) Velocity time courses were evaluated for PC data in the ascending aorta. Furthermore, a correlation analysis of pixel-wise velocity values for the peak velocity frame of the segmented ascending aorta was performed between the fully acquired k-space reconstruction and kt-SENSE / kt-GRAPPA. Results: kt-SENSE and kt-GRAPPA in Fig.1 show similar image quality compared to the full k-space reconstruction. The kt-SENSE noise image in Fig.2 demonstrates distinct noise amplification in the myocardial region with motion while static tissue regions exhibit considerably less noise. In contrast, ktGRAPPA shows only minor noise amplification in the myocardial region but increased noise in the static regions compared to kt-SENSE. These finding are also supported by the calculated SNR in the two ROIs (Table 1). Phase difference images of kt-SENSE and kt-GRAPPA in Fig.3 show similar artifact behavior if compared to the reference (see arrows). Fig. 4 and 5 demonstrate the excellent quantitative agreement of both acceleration techniques compared to the fully acquired k-space. Regional RMSE in the LV of ktGRAPPA was slightly superior compared to kt-SENSE due to less noise in this region.