INTRODUCTION

MR elastography has been recognized as a useful tool in studies of neural degenerative disease in the brain^1-3. However, resolving accurate tissue mechanical property from in vivo MRE scans requires high-spatial-resolution, full vector field encoding, multiple MRE phases and high signal-to-noise-ratio (SNR), which usually results in long acquisition time. In this regards, spiral acquisition has been developed in spin echo sequence with multislice/multislab schemes^4-5, providing improved tradeoff between resolution, SNR and scan time. Nevertheless, spiral acquisition in gradient echo (GRE) MRE has not been fully developed due to lower SNR. In this study, a novel 3D multishot, variable density spiral staircase (SSC) acquisition is proposed for GRE-MRE. We achieve high resolution by using improved through-plane parallel imaging inherent from the proposed SSC trajectory with much less SNR penalty and further boost the displacement-SNR by exploiting a MEG of 2 wave-period duration. In vivo human experiments have demonstrated that the proposed method could provide high-SNR shear wave and stiffness maps comparable to a standard single-shot spin-echo echo planar imaging scan and high resolution (e.g., 1.8×1.8×1.8mm³) whole-brain MRE is possible within 5 minutes.

METHODS

1) Data Acquisition: The proposed multishot, variable density 3D SSC trajectory is illustrated in Fig. 1. SSC uniformly distributes M groups of N spiral arms (each rotates by n*360°/N) along kz axis. Each set of M spiral arms (all have the same rotation about kz) is shifted by n*Δkz/N (Δkz is the Nyquist distance), resulting in a linear phase variation as a function of slice location. One advantage of this feature is that it introduces additional incoherence among image slices, leading to improved through-plane parallel imaging with much less SNR penalty.
The pulse sequence for the proposed method is shown in Fig. 2. To enhance the displacement SNR, a motion encoding gradient (MEG) of a duration of 2 wave period (T) is adopted. The multishot, variable density spiral-out readout follows next, achieving a short TE of 35 ms approximately optimal for phase imaging in the brain at 3T⁶. Note that the proposed technique provides flexibility in trade-off among resolution, SNR, spatial-blurring and scan time. Given a desired resolution, acquisition parameters such as through/in-plane acceleration factor, spiral readout time, variable density can be adjusted accordingly to meet a feasible scan time with sufficient SNR.
2) Image Reconstruction: After obtaining the under sampled SSC data, spiral images from each MRE phase and MEG direction were reconstructed by solving the following optimization problem:
$$\hat{\rho}=\underset{\rho}{\operatorname{argmin}} \sum_{l=1}^{L}\left \| G_{xy}^{H}FP_zS_l\rho - d_l \right \|^2_2 + \lambda R(\rho)$$
where $$$\rho$$$ denotes the target image function, $$$S_l$$$ the $$$l$$$-th coil sensitivity profile, $$$P_z$$$ the linear phase compensation term for SSC along z, $$$F$$$ the Fourier transform operator, $$$G_{xy}$$$ the 2D gridding operator and $$$H$$$ the Hermitian conjugate operator. $$$d_l$$$ contains the spiral data from the $$$l$$$-th coil. $$$R(·)$$$ is a regularization function (e.g., Tikhonov regularization) which allows to incorporate additional prior information. After reconstruction, spiral images were processed using a conjugate gradient deblurring method⁷ with a known $$$B_0$$$ map to correct the off-resonance induced blurring. Both the sensitivity and $$$B_0$$$ maps were obtained from a low-resolution prescan.
3) MRE inversion: Given the deblurred images, the motion phase component along each axis can be calculated according to the motion encoding directions. With the curl of the motion phase determined⁸, the shear modulus is reconstructed via direct inversion of the Helmholtz equation⁹.

RESULTS

In vivo experiments were performed on a 3T Phillips system (Ingenia Elition X) with a 14-channel receiver head coil on two healthy subjects in accordance with the institutional IRB and written informed consent. Brain MRE data at two different spatial resolutions were acquired using the proposed method (FOV 240×240×120 mm³, TR/TE 67/35 ms, FA 23°, 2× undersampling along kz, wave frequency 60Hz, 4 MRE phases, 6 MEG directions) in 1.5 and 4.5 minutes respectively ( 16 ms spiral-readout, 2 spiral-shots, 1.4× slice oversampling for 3×3×3 mm³ and 18.6 ms spiral-readout, 4 spiral-shots, 1.2× slice oversampling for 1.8×1.8×1.8 mm³). To validate the stiffness images from the proposed method, a standard single-shot 2D spin-echo EPI MRE sequence was also conducted at 3×3×3 mm³ resolution for comparison (TR/TE 4000/67ms, FA 90°, scan time 96 seconds).
MRE results generated from both methods were compared in Fig. 3. The proposed 3D gradient-echo acquisition provides high-SNR shear wave and stiffness images comparable to those from the 2D SE-EPI scan, while the conventional 2D scan will suffer from phase jitters across slices. The high-resolution results from another subject using the proposed method are shown in Fig. 4 with clearly identified brain regions (e.g., ventricles) in the stiffness maps, demonstrating the capability of the proposed method for high-quality MRE of the whole-brain.

CONCLUSION

A novel 3D GRE multi shot, variable density spiral staircase acquisition is proposed for whole-brain MRE. This technique integrates a 3D SSC trajectory, parallel imaging and a new MEG to achieve fast and high quality in vivo mapping of tissue mechanical property in the brain. Future work may include further optimizing scan parameters to provide better trade-off between SNR, resolution and scan time.

Acknowledgements

This work is supported in part by a research funding from Philips.