Evaluation of through-time radial GRAPPA for real-time cardiac MR imaging at 7 Tesla
Sascha Brunheim1,2, Sören Johst1, Stefan Maderwald1, Stefan Rietsch1,2, Stephan Orzada1, Marcel Gratz1,2, Juliane Goebel3, Kai Nassenstein3, Nicole Seiberlich4, and Harald H. Quick1,2

1Erwin L. Hahn Institute for Magentic Resonance Imaging, University Duisburg-Essen, Essen, Germany, 2High Field and Hybrid MR Imaging, University Hospital Essen, Essen, Germany, 3Department of Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany, 4Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States

Synopsis

Accelerated radial data acquisition of the myocardium in combination with through-time radial GRAPPA offers the opportunity for real-time visualization of cardiac motility without the need for additional ECG or pulse wave synchronization. This is particularly useful in an ultrahigh-field MR environment where conventional gating methods in combination with cardiac dysrhythmia tend to fail. In this work, the performance of through-time radial GRAPPA against an established Cartesian k-space encoding protocol featuring pulse-triggered cine-FLASH has been evaluated and its role as an alternative for real-time cardiac 7-Tesla MR imaging is shown.

Purpose

The use of real-time, free breathing, non-gated cardiac imaging (rtCMR) is desirable for assessment of myocardial contractility when using ultrahigh field strengths, as the magneto-hydrodynamic effect[1] at 7T MRI limits the reliability of ECG. This, in combination with arrhythmias or the inability to comply with breath-holding, renders the acquisition of segmented cardiac cine images at 7T difficult. Therefore highly undersampled non-Cartesian trajectories with a through-time radial (TTr) GRAPPA reconstruction[2] were evaluated for rtCMR. A frame rate of 20 Hz together with a clinically relevant spatial resolution without significant motion or aliasing artifacts is necessary for this purpose. Cartesian, pulse-triggered, breath hold, segmented cine-FLASH cardiac MRI served as standard of reference.

Methods

Data were acquired on a 7T research system (Magnetom 7T, Siemens Healthcare GmbH, Germany) equipped with an 8ch Tx/Rx flexible body coil[3] and a custom SAR supervision system[4]. Measurement preparation included B0 shimming and multi-channel RF shimming. Short axis views and horizontal long axis views of myocardium were acquired in N=5 healthy subjects, employing four different accelerated radial fast low flip angle gradient-echo (FLASH) protocols with parameters as shown in tabular form in Figure 1. These rtCMR protocols were adjusted for optimal spatial and temporal resolution with an acceleration factor of R=8, and were set such that the field-of-view extended only around the heart itself. In addition to the accelerated radial data, a separate fully-sampled calibration dataset was also collected for through-time radial GRAPPA reconstruction. The radial data underwent an off-line post-processing procedure[2] with MATLAB (The MathWorks Inc., Natick, USA) testing different numbers of calibration frames and different radial GRAPPA segment sizes within read-out (kr) and projection (kθ) direction (4x1 and 8x1 with 112 calibration frames; 8x4 and 4x4 with 28 frames) for TTrGRAPPA reconstruction (Figure 2). For comparison, a segmented cine-FLASH sequence with a Cartesian trajectory and retrospective pulse triggering within breath-hold at same slice position served as standard of reference. Contrast ratios in comparable regions-of-interest between blood and myocardium (CRblood/myo) were calculated for a full cardiac cycle of the short axis views for both sequence types.

Results

The through-time radial GRAPPA with a segment size of 8x1 and 112 calibration frames showed the best performance in terms of CRblood/myo. With this large number of calibration frames, the total imaging time for calibration/undersampled data acquisition was at least twice the time needed for the retrospective pulse triggered cine-FLASH sequence for the healthy subjects without dysrhythmia. However the cine-FLASH depended on a correct gating and is susceptible to movement other than that of the myocardium (e.g. breathing). Figure 3 shows the streak artifact free image quality of selected T1-weighted frames for the real-time method and the corresponding image portions of the pulse gated cine sequences with the same spatial resolution. In Figures 3 and 4, the blood pool appears to be more inhomogenous in the highly accelerated (R=8) real-time protocols in contrast to the cine-FLASH with flow compensation. Furthermore, the CRblood/myo for one cardiac cycle of short axis view was roughly the same for triggered and real-time sequences acquired with identical voxel size. When increasing the spatial resolution of the real-time measurement by 50% to 1 mm in-plane and using a lower temporal resolution of 17 fps, the mean CRblood/myo was reduced by 63% (Figure 4). Mainly this high in-plane resolution was chosen to test the limits and is normally not necessary for rtCMR.

Discussion

TTrGRAPPA may serve as an alternative to the established cine technique at 7T because it works independently any gating modality and the patient compliance to breathing commands. Nevertheless, a direct comparison between both techniques is not trivial due to the different breathing conditions and the selection of exactly identical time frames. One drawback of the real-time method is the necessity for acquisition of a calibration dataset with a sufficient number of spokes recorded repeatedly to satisfactory determine the GRAPPA weights and the associated resulting prolonged scan time. However, other groups have demonstrated that whole ventricular coverage with the through-time radial GRAPPA technique can be performed in approximately half the scan time of a standard cine by reducing the number of calibration frames[5]. Despite the fact that our rtCMR measurements took more time than the cine-FLASH it shows the opportunity to record the heart/blood movement for the complete gating-free imaging process while free breathing. In this study, the acceleration factor and thus temporal and spatial resolution was limited by the hardware available. An additional evaluation with an 8ch Tx/32ch Rx flexible body RF coil is in progress whereas acceleration factors of more than R=8 are reasonable.

Acknowledgements

No acknowledgement found.

References

[1] Schenck JF “Physical interactions of static magnetic fields with living tissues.” (2005) Prog Biophys Mol Biol 87(2-3): 185-204. [2] Seiberlich N, Ehses P, Duerk J et al. “Improved radial GRAPPA calibration for real-time free-breathing cardiac imaging.” (2011). MRM 65(2): 492-505. [3] Orzada S, Quick HH, Ladd ME et al. “A flexible 8-channel transmit/receive body coil for 7 T human imaging” (2009) Proc. Intl. Soc. Mag. Reson. Med. 17: p. 2999. [4] Bitz AK, Brote I, Orzada S et al. “An 8-channel add-on RF shimming system for whole-body 7 tesla MRI including real- time SAR monitoring.” (2009) Proc. Intl. Soc. Mag. Reson. Med. 17: p. 4767. [5] Aandal G, Nadig V, Yeh V et al. “Evaluation of left ventricular ejection fraction using through-time radial GRAPPA” (2014) JCMR 16(1): 79.

Figures

Fig.1 Experimental parameters for the non-triggered, real-time radial TTrGRAPPA and for the pulse-triggered, breath-hold Cartesian cine-FLASH sequence.

Fig.2 Schematic of TTrGRAPPA acquisition with factor R=2 undersampled accelerated radial data and certain repetitions (= number of frames) for determination of the 2x3 GRAPPA kernel. Sample images (R=8) before TTrGRAPPA reconstruction without aliasing but severe streaking artifacts per acquisition of frame.

Fig.3 Qualitative comparison of short axis views and horizontal long axis views during the mid-diastolic cardiac phase of one subject at the same slice position (overall resolution 1.4x1.4x5mm3). A+C: Standard segmented cine-FLASH images in breath-hold. B+D: TTrGRAPPA reconstructed radial FLASH images (R=8) while free breathing.

Fig.4 One complete short axis view cardiac phase of one subject with calculated CR between left-ventricular blood and myocardium signal. A: TTrGRAPPA (R=8) time series with 17fps and voxel size 1x1x5mm3. B: TTrGRAPPA (R=8) time series with 20fps and voxel size 1.4x1.4x5mm3. C: Segmented cine-FLASH time series with voxel size 1.4x1.4x5mm3.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3527