Introduction

MR Elastography (MRE) encodes the harmonic motion of tissue into the MR phase signal (1). However, 2π phase wrapping limits the range of the motion that MRE can reliably measure. Though existing image processing algorithms can typically undo spatial wrapping, unwrapping in the temporal dimension with only 4-8 samples (the typical number of phase offsets in MRE experiments) has remained very challenging (2,3). In an application where both small shear-wave (1-10 μm) and large rigid-body (10-50 μm) motions are of interest, the required large dynamic range poses significant challenges for conventional MRE (3). One can perform two separate MRE scans using different motion-encoding gradient (MEG) amplitudes (3) to achieve high and low motion sensitivity. However, this increases the scan time and sacrifices encoding efficiency. In this study, we developed a novel framework that simultaneously acquires dual-motion-sensitivity MRE data and beneficially exploits data redundancies during processing. This framework yields motion measurements whose dynamic range is increased 8-10 times compared to conventional MRE, and enables both shear-wave and large bulk (rigid-body) motion to be characterized in a single scan.

Theory

In a standard MRE acquisition with monodirectional encoding, 2 acquisitions are performed with the same MEG amplitude but opposite polarity, and phase-difference images Φ=θ₊-θ_-are calculated to remove any unwanted background phase and double the motion sensitivity. These images can exhibit substantial phase wraps. In the proposed scheme (Fig. 1), the negative MEG is set to be smaller than the positive MEG. The sum of the two phase images φ=θ₊+θ_- generates a low-sensitivity phase image (no phase wraps) that can be used to estimate and guide the unwrapping of Φ. In practice, the mandatory z-axis, flow-compensation (FC) gradients encode some amount of motion into the low-sensitivity encoded phase φ_x and φ_y, resulting in large estimation errors if not corrected. To address this problem, we set ±z MEGs to be equal, but use the z-axis FC gradients to provide the low-motion-sensitivity. The φ_x and φ_y images can then be corrected by subtracting out φ_z . The phase estimation was performed on the first temporal harmonic of the phase offsets, and then projected to each phase offset for unwrapping (Fig. 2).

Method

MRE scans were conducted on 3T GE scanners with an 8-channel head coil. A phantom study was performed using 60-Hz vibrations. The negative MEGs were set at 77.7% of the positive MEGs, except for the z MEGs which used the same amplitude for both polarities, resulting in an 8-times ((1+0.777)/(1-0.777)) difference in the motion sensitivity. For validation, low-sensitivity phase images were also acquired from a separate standard MRE scan using small MEGs. The same MEG setup was used for a volunteer study with the following imaging parameters: TR/TE=4000/58.7 ms; FOV=24 cm; 80×80 acquisition matrix reconstructed to 128×128; 48 contiguous 3-mm-thick axial slices; 2×ASSET acceleration; 4 phase offsets, and 3 minutes acquisition time.

Results

The phantom results (Fig. 3) showed that both the wrapped high-sensitivity phase images (Fig. 3a) and its wrap-free estimate (Fig. 3b) generated from the low-sensitivity phase images were acquired simultaneously in a single scan. Fig. 3c shows the unwrapped high-sensitivity phase images. The estimated phase closely resembled the unwrapped phase and the reference images from the separate low-sensitivity scan (Fig. 3d). The volunteer results in the axial (Fig. 4) and sagittal (Fig. 5) views demonstrated that the phase wraps can be simply and efficiently eliminated throughout the whole 3D volume. Note that in Fig. 4d, the large displacement at the example voxel was correctly sampled by the low-sensitivity encoding, and was then recovered from the high-sensitivity wrapped phase, which would be very difficult using conventional unwrapping algorithms. The large harmonic motion (7.2π from peak to peak; 3 phase wraps between offsets 1 and 2) can be resolved using as few as 4 phase offsets, and the smaller shear wave (<0.4π, Fig. 4e) was simultaneously detectable with a reasonable SNR.

Discussion

The dynamic range of accurate MRE displacement information is significantly increased by the proposed scheme, at the expense of slightly (10-20%) reduced encoding efficiency due to the smaller negative MEG amplitude. The low motion sensitivity in the z-direction currently depends on the slice thickness because the FC gradient changes based on the properties of the slice-selection gradient. In the future, we will further optimize this approach by allowing a small portion of the z-MEG to be always on to compensate for this change.

Conclusion

We have developed a dual-sensitivity motion-encoding scheme and demonstrated that it has the capability to increase the dynamic range of the motion that MRE can reliably detect.

Acknowledgements

This work was supported by the grant from the National Institute of Health RO1 EB001981.