Highly-efficient free breathing whole heart CINE MRI with self gated 3D CASPR-TIGER trajectory
Muhammad Usman1, Gastao Cruz1, and Claudia Prieto1

1Division of Imaging Sciences and Biomedical Engineering, King's College London, London, United Kingdom

Synopsis

In this work, we propose to use a novel free-running self-gated 3D CArtesian acquisition with Spiral PRofile ordering and TIny Golden angle step for Eddy Current Reduction, so called CASPR-TIGER. Data is acquired continuously under free breathing (no ECG gating, no pre-pulses interruption) using CASPR-TIGER trajectory and 4D volumes (3D+time) are reconstructed from all available data (100% respiratory scan efficiency) using a soft gating technique combined with temporal total variational (TV) constrained iterative SENSE reconstruction. Feasibility of proposed method is demonstrated in three subjects in a 3-3.6 minutes free breathing acquisition.

Introduction

3D whole heart CINE MRI has the advantage of large spatial coverage, no slice-misalignments and less rigorous planning requirements compared to conventional multi-slice breath-hold 2D acquisitions. However, respiratory gating required in 3D CINE MRI leads to prolonged scan time. Recently, 3D radial self-gated approaches have been proposed to accelerate the acquisition, without1,2 and with3 the use of blood-pool contrast agents. However, these techniques are computationally very complex due to 3D gridding/re-gridding operations. In this work, we propose to use a novel free-running self-gated 3D CArtesian acquisition with Spiral PRofile ordering and TIny Golden angle step for Eddy current Reduction, so called CASPR-TIGER. Data is acquired continuously under free breathing (no ECG gating, no pre-pulses interruption) using CASPR-TIGER trajectory and 4D volumes (3D+time) are reconstructed from all available data (100% respiratory scan efficiency) using a soft gating technique combined with temporal total variational (TV) constrained iterative SENSE reconstruction. Feasibility of proposed method is demonstrated in three subjects in a 3-3.6 minutes free breathing acquisition.

Method

Trajectories with golden angle step are advantageous for free-running acquisitions as these allow retrospective adjustment of temporal resolution by combination of any arbitrary number of profiles. However, it has been shown that in combination with balanced SSFP sequence, the large azimuthal angle increment can lead to strong image artefacts due to eddy currents in 2D golden radial acquisitions. Recently, the use of tiny golden angle (23.628°) has been proposed to reduce these artefacts4. 3D CASPR-TIGER trajectory acquires data along spiral like interleaves in phase encoding ky-kz plane, with tiny golden angle step between consecutive spiral interleaves. To further avoid artefacts that can result from high to low frequencies jump in k-space, the interleaves are acquired in a paired fashion (Fig.1). The first interleaf in each pair starts at the outside part of k-space and ends in the centre part of k-space. The second interleaf starts in the centre k-space and ends in the outside part of k-space. For the purpose of self-respiratory navigation, the k-space centre is acquired at the beginning of second interleaf in each pair. After acquisition of a number of interleaves that populate full k-space, a shift in the initial angle is introduced to avoid overlapped samples in the final reconstruction. The R-wave of the ECG is logged for the purpose of retrospective cardiac synchronization. The respiratory signal is estimated from repeatedly acquired 1D projections and is used to define a reference bin at the end-expiration. Soft gating5 is performed on the acquired data to weight k-space data depending on displacement from the reference bin. 4D temporal total variational constrained iterative SENSE (TV-SENSE)6,7 reconstruction is done on the soft-gated data with TV applied along the cardiac phase dimension.

The reconstruction process can be formulated as:

$$$ arg min_{\bf{x}}\parallel{\bf{W(Ax-y)}\parallel}^2_2 +\lambda_t\|\triangledown_t\bf{x}\|_1 \space\space\space\space\space Eq[1]$$$

where $$$\bf{y}$$$ is the acquired data, $$$\bf{x}$$$ is the reconstructed 4D volume (3D+time), $$$\bf{A}$$$ is the encoding operator that includes the coil sensitivities, Fourier transformation and sampling, $$$\triangledown_t$$$ represents 1D temporal gradient, $$$\lambda_t$$$ is a scalar used to define balance between data consistency and TV regularization, $$$\bf{W}$$$ weights each readout according to its displacement from the reference bin using a scaled Gaussian kernel with the maximum set to 1.

Experiments

Whole-heart free breathing acquisition was performed with CASPR-TIGER trajectory on a 1.5T scanner (Ingenia, Philips Healthcare) in three subjects using a b-SSFP sequence (TR/TE=3.2/1.6 ms, resolution = 2 mm3 isotropic, FOV = 320 x 320 x 90-100 mm3, number of spiral interleaves = 4152-4952, 14 readouts per spiral interleaf, scan time= 3-3.6 mins). A reference bin with width of 4mm was defined at end expiration for soft-gating. Sixteen cardiac phases were retrospectively reconstructed using soft-gated 4D TV-SENSE reconstruction in Eq[1]. 4D volumes were reformatted in different planes after reconstruction.

Results and Discussion

The 1D projection from centre k-space profiles and corresponding respiratory signal are shown in Fig2a. Reconstructed images for 2-chamber, 4-chamber and short axis planes in diastolic and systolic phases are shown in Fig.2b for one subject. Different slices in short axis orientation from 4D reconstructed volume are shown in Fig.3 for a different subject. The reconstructed images using proposed method had good contrast and spatio-temporal variations, correctly recovering diastolic and systolic cardiac phases. Future work will investigate quantitative evaluation of left ventricle functional parameters using the proposed method and compare obtained values with those provided by gold standard 2D multislice breath-hold acquisition.

Acknowledgements

The work presented in this paper is funded by Medical Research Foundation (MRF) grant MRF-C0510.

References

1. Coppo, S., Piccini, D, Bonanno, G, Chaptinel, J, Vincenti, G, Feliciano, H, van Heeswijk, R., Schwitter, J and Stuber, M (2015), Free-running 4D whole-heart self-navigated golden angle MRI: Initial results. Magn Reson Med, 74: 1306–1316. doi: 10.1002/mrm.25523

2. Feng L, Coppo S, Piccini D, Lim RP, Stuber M, Sodickson DK, and Otazo R, Five-Dimensional Cardiac and Respiratory Motion-Resolved Whole-Heart MRI, P27, ISMRM 2015.

3. Pang J, Sharif B, Fan Z, Bi X, Arsanjani R, Berman DS, Li D. ECG and navigator-free four-dimensional whole-heart coronary MRA for simultaneous visualization of cardiac anatomy and function. Magn Reson Med 2014;72:1208–1217

4. Wundrak S, Paul J, Ulrici J, Hell E, Geibel MA, Bernhardt P, Rottbauer W, Rasche V. Golden ratio sparse MRI using tiny golden angles. Magn Reson Med, doi:10.1002/mrm.25831.

5. Forman C, Piccini D, Grimm R, Hutter J, Hornegger J, Zenge MO. Reduction of respiratory motion artifacts for free-breathing wholeheart coronary MRA by weighted iterative reconstruction. Magn Reson Med 2014;73:1885–1895

6. Otazo R, Kim D, Axel L, Sodickson D. Combination of compressed sensing and parallel imaging for highly accelerated first-pass cardiac perfusion MRI. Magn Reson Med 2010;64:767–776

7. Cruz, G., Atkinson, D., Buerger, C., Schaeffter, T. and Prieto, C. (2015), Accelerated motion corrected three-dimensional abdominal MRI using total variation regularized SENSE reconstruction. Magn Reson Med. doi: 10.1002/mrm.25708

Figures

Fig1: Free-running CASPR-TIGER: First interleaf (blue) is acquired from outer to central k-space region, second interleaf (green) is acquired in opposite direction. k-space centre profile is acquired at each interleave pair (red). Each interleaf is at 23.62o with respect to previous one to ensure uniform k-space coverage and reduced eddy current artefacts.

Fig2: Self-gated 3D whole heart CINE MRI: a) 1D projection obtained from self-navigation and corresponding estimated 1D respiratory signal, b) Diastolic and systolic phases in two, four chamber and short axis orientations for subject one.

Fig3: Self-gated 3D whole heart CINE MRI using proposed framework: short axis slices from base towards apex are shown in diastole (top row) and systole (bottom row) for subject two.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1814