Rapid cardiac cine imaging using MACH

Introduction

Previously, we developed a sparsely distribute k-space-time sampling approach termed BRISK, Block Regional Interpolation Scheme for K-space [1]. This approach allowed a nominal acceleration factor of 4 with good quality and low artifact. Others have developed alternative k-space-time sampling schemes, such as KT-BLAST and SLAM [2, 3]. We note that even at high acceleration factors, KT-BLAST/SLAM allowed a smooth transition from frame to frame, while BRISK experienced ringing artifacts. From these considerations we isolated key features that contribute to a successful k-space-time sparse sampling scheme:

1) Update of k-space should be rapid near the center and lower near the periphery (as in BRISK) to capture highly dynamic features.

2) Update of k-space should smoothly vary over time (as in KT-BLAST/SLAM) avoiding sudden transitions between k-space regions to result in smoother transition between frames.

From these design principles we developed a new sparse k-space-time sampling scheme, MACH, Multiple Acceleration Cycle Hierarchy. MACH incorporates a gradually changing rate over time, starting with the highest rate near the center of k-space and becoming progressively sparser towards the periphery, Figure 1 shows the k-space-time sampling scheme. In MACH, the progressively decreasing sampling rates are not confined to integer steps, thereby providing greater opportunity for a smooth transition over the k-space-time domain. Data that are not directly sampled in MACH are interpolated prior to applying conventional Fourier transformation to generate the series of images. Since each frame incorporates a full set of k-space data, the SNR is similar to the conventional scan.

Figure 1

Figure 1

Methods

Simulations were performed using fully acquired stead-state-free-precession (SSFP) cine image data to allow direct comparison of MACH and KT-BLAST/SLAM when using the same acceleration factors. Further, MACH was implemented on a 1.5 T scanner (GE, Milwaukee, WI). Using the SSFP cine acquisition, long and short axis acquisitions were acquired using the conventional examination and MACH applied with a net sparse factors ranging from 2 to 5, with matrices ranging from 224 to 336. Comparable cine acquisitions were acquired of the heart in 10 volunteers. The end-diastolic and end-systolic phases were identified and areas were planimetered and compared between the conventional and the MACH accelerated scans.

Results

Simulations showed that for a moderate acceleration factor of 5, KT-BLAST/SLAM represented myocardial motion well, but lost detail of the relatively fast moving valvular features, while MACH still represented these features. In the in vivo acquisitions, the average MACH acceleration factor applied was 3.5 ± 1.1, end-diastolic and end-systolic ventricular chamber areas were not significantly different between the conventional and the accelerated MACH scans (p = 0.7, 0.6, respectively) and correlations were excellent at 0.99 for each. Compared to the conventional scan, there is no additional overhead with MACH. See Figures 2 and 3.

Figure 2

Figure 2

Figure 3

Figure 3

Conclusion

In sparse k-space-time acquisition strategies, rapidly updating the central region of k-space is known to be important. We note that MACH achieves this condition very efficiently while also achieving a smooth transition of update rate between each region of k-space since MACH does not use a uniform or even a regular update rate. MACH was successfully implemented and shown to accurately represent cardiac regions with good fidelity.