How MUSE is changing diffusion MRI

When coupling MUSE with the ultra-high-powered prototype gradient coil‡ in the SIGNA™ Premier 3.0T system at Duke University, the spatial resolution limits of diffusion data can be pushed even further, with whole brain DWI images acquired with 0.7 mm isotropic voxels (Figure 2). Although highly beneficial in brain imaging, MUSE’s ability to account for motion artifacts in sub-millimeter diffusion images is equally significant in body imaging. Figure 3 compares a simple MUSE DWI abdominal scan using 1 mm³ isotropic voxels to that of 0.6 x 0.6 x 2.0 mm, where the anatomical specificity in both the spinal cord and kidneys is immediately apparent in the sub-millimeter scan.

Beyond the increase in achievable spatial resolutions, one of the most appealing aspects of the MUSE framework is the foundation it provides for incorporating additional imaging tools and techniques. For example, the traditional 2D MUSE model has been expanded to acquire and reconstruct 3D ms-EPI data, where an additional SNR advantage is gained by Fourier encoding thick slabs across the imaging volume to produce images with increased anatomical detail (Figure 4).‡

Although the shortened readout and echo spacing of ms-EPI does reduce distortion artifacts in DWI, a recent adaptation of MUSE (marked by GE as PROGRES) to incorporate reversed polarity gradients (RPG) between shots has shown great promise in eliminating both residual static field inhomogeneities as well as dynamic gradient-induced eddy currents. While traditional RPG distortion corrections require additional imaging volumes with opposing phase encoding directions (PE), RPG-MUSE simply reverses PE in the odd/even shots of a typical ms-EPI scan. This approach requires no additional volumes or increase in scan duration, as it allows for field inhomogeneities to be estimated directly from the ms-EPI data and incorporated into the reconstruction. As shown in Figure 5, the resulting RPG-MUSE images exhibit tremendous anatomical specificity when compared with a traditional correction of data with a single PE.

With its small cross-sectional area, another region that has benefited greatly from the high resolution afforded by MUSE is in the spinal cord. Unlike traditional ms-EPI, where each shot is separated by the duration of a predefined TR, spinal cord imaging often employs cardiac gating (CG) in an effort to mitigate cerebrospinal fluid pulsation artifacts. Although CG both increases and varies the time between shots, MUSE remains fully capable of accounting for intershot motion artifacts. As illustrated in Figure 6, the sub-millimeter spatial resolution and SNR achieved by MUSE facilitates the delineation of complex microstructures in the brain stem, such as the bifurcation of corticospinal tracts and the decussation of medullary pyramids.

Despite its remarkable advantages, the one remaining drawback of ms-EPI is the prolonged acquisition time. To that end, the MUSE framework has since been expanded to incorporate simultaneous multi-slice imaging, or HyperBand. Through HyperBand, a whole brain 1 mm³ MUSE scan with 4 shots and 25 diffusion volumes can be reduced from approximately 20 minutes to around 7 minutes, while preserving both the increased SNR and spatial fidelity of ms EPI (Figure 7). Alternatively, MUSE with HyperBand can also be used to extend the coverage achieved within a given scan time. This is of particular use in applications such as spinal cord imaging, as it would enable fiber bundles to be tracked at sub-millimeter isotropic spatial resolutions from their origins within the brain as they extend down into the spinal cord.