Diffusion-weighted Imaging of the Spinal Cord: Is There a Future?

MR imaging of the brain has benefited over the past few years by the routine use of a number of innovative pulse sequences, prominent among which has been diffusion-weighted imaging (DWI). The efficacy of this technique, well known to virtually all radiologists, has found its greatest use in the evaluation of cerebral ischemia, although application to other brain abnormalities has been shown in numerous publications. The incremental information that DWI brings to brain imaging is formidable, so naturally the desire would be to add DWI to the MR study of a patient with spinal cord dysfunction. One could envision many situations in which DWI would be extremely helpful, both in terms of diagnosis and evaluation of the efficacy of medical and surgical treatment. Clearly, spinal cord infarcts resulting from either arterial or venous abnormalities come to mind first, not just because their cerebral counterparts can be identified so well, but because cord infarct/ischemia is frequently a difficult diagnosis to make on the basis of MR findings. Other entities, such as acute transverse myelopathy, Wallerian degeneration, or acute disseminated encephalomyelitis might be diagnosed earlier and characterized better if proper technical parameters for obtaining spinal cord DWI and calculations of apparent diffusion coefficients could be achieved. Unfortunately, there are many technical and physiologic problems to overcome before DWI of the spinal cord becomes an accepted and routinely used protocol.

In this issue of the AJNR, Robertson et al (page 1344) describe their work on line-scan diffusion imaging of the spinal cord in 12 children, and in three cases they compare line-diffusion scanning with echo-planar diffusion imaging (EPDI). As one might expect, line scanning resulted in a better signal-to-noise ratio, and there were diminished magnetic susceptibility effects. The wide range of relative anisotropy in the normal cords is explained by both the curved nature of the spinal cord relative to the orthogonal diffusion-gradient axes, as well as the averaging of gray and white matter in their measurements. With this article in mind, and with an increased interest in cord DWI, it is clear that a number of problems must be successfully dealt with before DWI of the cord becomes part of routine spine imaging. The inherent difficulties in obtaining high-quality DWI of the cord, and techniques for dealing with these problems, deserve comment.

A major difficulty in obtaining DWI of the cord is physiologic motion, particularly CSF flow, which causes imaging artifacts. A number of strategies can be employed to overcome the artifacts that pulsating CSF generates, including novel means of data acquisition such as the navigator-echo or line-scanning technique, fast imaging methods such as single-shot EPDI or single-shot fast spin-echo (FSE) imaging, and cardiac gating. Problems exist, however, with each of these strategies. Specifically, when cardiac gating is used, the examination time is extended because only a limited number of images are acquired when cord motion is minimal (ie, during diastole). When the navigator-echo method with fast imaging or line scanning is used, lower signal is obtained. Finally, when EPDI is used, magnetic susceptibility artifacts and low spatial resolution of this relatively small structure result in suboptimal image quality.

The navigator-echo method, which uses an extra spin-echo sequence with no spatial phase encoding, provides phase shift information due to bulk motion, and these data are used to correct for phase shifts before the images are reconstructed. In addition, when the navigator-echo technique is used in conjunction with cardiac gating, diffusion-weighted images can be acquired throughout the entire cardiac cycle, not just during minimal motion, and this reduces scan time. Despite the advantages of the navigator-echo method and cardiac gating, the fast scanning techniques that are used in conjunction with them come with certain drawbacks. Specifically, EPDI, which uses multiple gradient echoes to acquire data, suffers from local susceptibility artifacts. This problem can be overcome by the use of a single-shot FSE sequence, which uses 180° refocusing pulses, rendering it less sensitive to artifacts caused by local magnetic field variations. With both single-shot FSE and EPDI, however, broad receiver bandwidths are used, which diminish the signal-to-noise ratio.

Line scanning differs in a number of ways from the conventional 2D Fourier transform imaging methods. This spin-echo–based technique acquires data from individually excited columns (or lines), and because no phase-encoding gradient is needed, artifacts due to physiologic motions are minimized. In addition, because line scanning uses a spin-echo rather than a gradient-based sequence (EPDI), susceptibility artifacts are minimized. Despite these advantages, an adequate signal-to-noise ratio is a problem in line scanning because data are received from just one column (line) of tissue rather than from an entire slice of tissue.

When one considers the issue of spatial resolution and the desire to image a patient's spinal cord with a resolution that approaches in vitro cord imaging, the problem of insufficient signal is clear. It may be that, with 1.5-T scanners and the current generation of receiver coils, none of the above-mentioned techniques offers the high signal and resulting image quality that would allow DWI of the cord to be widely and routinely implemented.

The answer to this dilemma of adequate signal combined with reasonable spatial resolution in DWI may require a rethinking of our approach to “physiologically based” images of the spinal cord. Specifically, in addition to innovative acquisition techniques or faster scanning, high field strength systems (3T or higher), different types of receiver cord design, or both eventually may be a successful approach. More signal from such system redesign may be the approach needed to give DWI of the spinal cord a future.