The purpose of this work is to develop a 3D MR Elastography pulse sequence that minimizes the acquisition samples needed for 3 displacement polarizations while also minimizing image artifacts. The complex wave fields sampled for Elastography contain 3 unknowns in the acquired phase. There is a static term and an in-phase and out-of-phase term for the wave field. These 3 unknowns require 3 samples for each of the 3 displacement fields, thus a total of 9 samples are needed. This approaches the 8 samples used for tetrahedral acquisition schemes and, more recently, SLIM-MRE¹. Additionally, acquisitions using 2D SE-EPI (Spin Echo Echo Planar Imaging) are increasingly being used for 3D processing. The data from EPI scans have periodic slice-to-slice phase artifacts. These artifacts require filtering² and can reduce the apparent stiffness. This work uses a 3D GRE (Gradient Recalled Echo) to sample the displacement fields. Given the 3D nature of the acquisition, no filtering is needed.

Previously³, a 3D GRE sequence using 4 time points and 12 total samples was presented. This work extends this technique to 3 time points. The sequence does not invert the MEG (motion encoding gradients) gradients and instead uses the TR to advance the phase of the complex displacement field. The timing schedule of the acquisition is seen in figure 1. The image reconstruction of the 3 phase difference images follows the standard vascular reconstruction⁵ for a multi-coil data set and is schematically seen in figure 2. When a phase difference is calculated for time samples 180° apart, the phase values double. However, when the samples are separated by 120° the increase is $$$\sqrt{3}$$$. To preserve scaling of the phase to physical motion, this difference is noted in the reported MENC values. After phase unwrapping, a 3 point discrete Fourier transform can then be used to generate the complex displacement field. If a vector curl is applied during processing, a static phase term needs to be removed from the y and z encoded fields due to there relative shift in the complex plane. The constant $$$e^{-iN2\pi/9}$$$ is multiplied to the y encoded field, where N is the number of motions cycles covering the 9 acquired samples. Twice the phase is multiplied to the z encoded field.

Mayo Clinic IRB approved scans were obtained of head and breast anatomy. Scans were performed on a GE 3.0T scanner. The breast MRE parameters used a fat/water in-phase TE/TR= 18.176/22.2, mechanical vibration=40Hz, a fractional 1-2-1 MEG pulse for flow compensation, MENC=12.3μm/rad, FOV=380x380x160mm, acquisition size=96x96x40, resolution=4x4x4mm, flip angle=8, ARC phase reduction factor=2, scan time=5:39minutes. The results are seen in figure 3. The breast scans show ample wave sensitivity and image quality. Reducing the needed samples to 9 allows for full coverage in a reasonable scan time of 5:39.

The head 3D sagittal GRE scan parameters TE/TR= 28.4/35.2, mechanical vibration=60Hz, a 1.43 cycle MEG pulse that provides flow compensation, MENC=11.5μm/rad, FOV=220x220x180mm, acquisition size=74x74x60, resolution=3x3x3mm, flip angle=12, 2D ARC reduction=2x2, scan time=6:46minutes. The GRE head scan was compared to 2D axial EPI parameters of TE/TR= 62.3/3601, mechanical vibration=60Hz, two 1-2-1 MEG pulses, MENC=6.5μm/rad, FOV=240x240x144mm, acquisition size=74x74x48, resolution=3.2x3.2x3mm, ASSET reduction=3, scan time=6:47minutes. The comparison is seen in figure 4. In the head application the EPI scans show residual banding artifact in the sagittal reformatted images even after processing to remove them. These bands would appear as high frequency shear waves to inversion algorithms and result in reduced apparent stiffness estimates. The GRE scans show no slice-to-slice banding in the reformatted axial images. While the GRE scans showed reasonable penetration, the sensitivity to motion and SNR were limited.

Tetrahedral sampling strategies and SLIM-MRE can acquire all three displacement fields in 8 samples. However, as noted previously⁵, imaging and MEG gradients can cause table motion and induce shear waves in the body. The three repeating pulse sequence acquisition units, x-y-z, are only updated for phase encoding throughout the scan. This provides the same motion history, including induced table motion, and MEG induced eddy currents for all samples of a given polarization. This should be seen in a flatter divergence field compared to other minimum sample strategies.

The GRE sequence is the sequence of choice in our current breast studies. The head scans were performed using gradients with a maximum performance of 4 Gauss/cm. It would be desirable to use higher performance gradients with the GRE sequence. This would allow for both lower TE times and higher sensitivity to motion.