Joseph Helpern
Water molecules diffusing in the brain can be used to probe tissue microstructure noninvasively with diffusion MRI (dMRI). The microstructural distances to which dMRI is most sensitive are set by the typical distance a water molecule diffuses during the approximately 100 ms needed to acquire an individual dMRI signal. This distance of about 10 μm makes dMRI ideal for investigating cell-sized microstructural features. In this sense, dMRI overcomes the resolution limits of other MRI techniques and bestows upon dMRI a unique ability to noninvasively assess the microstructure of human brain tissue in vivo. This is a key reason behind dMRI’s rapid development into one of our premier imaging tools for understanding the brain.
One of the most popular and well-established dMRI methods is diffusion tensor imaging (DTI).1 However, the early approaches for many technologies, such as DTI, are usually followed by advances that strive to improve upon their performance and to overcome their limitations. In this regard, there are two critical issues for which DTI is limited. The first deals with crossing fibers within a voxel. Here, the tensor model is incapable of resolving multiple fiber orientations, which affects its reliability for fiber tractography. A second, broader issue is that DTI samples only a fraction of the information contained in the dynamics of molecular water diffusion. Interestingly, these two limitations are closely related to each other in that addressing the second leads to a resolution of the first.
Several approaches to solving the issue of crossing fibers have been implemented (see reference 2 for review)2 like high angular resolution diffusion imaging (HARDI)3 and the related Q-ball imaging.4 Diffusion spectrum imaging (DSI)5 is a more comprehensive approach that addresses both issues but at the expense of somewhat longer scan times. Diffusional kurtosis imaging (DKI) is a minimal extension of DTI that harvests additional information related to water diffusion by targeting the precise quantification of diffusional non-Gaussianity and does so within a clinically acceptable timeframe.6 The non-Gaussian nature of water diffusion in brain is believed to contain a rich store of information related to many aspects of tissue microstructure. Indeed, knowledge of diffusional non-Gaussianity has recently been demonstrated to provide a means of resolving crossing fibers.7,8
Additional advanced dMRI methods are on the horizon, like oscillating gradient9–12 and multiple-pulsed13–15 dMRI sequences. A recent and important observation for the special case of a double pulsed dMRI sequence has been made by Jespersen,16 who demonstrated that the unique information this method provides only appears in the non-Gaussian contributions to double-pulsed dMRI signal. This underscores the importance of incorporating the effects of non-Gaussian diffusion in dMRI measurements, and we have recently exploited this observation to introduce double-pulsed DKI.17 These new non-Gaussian, multiple-pulsed, and oscillating gradient dMRI methods can provide effective approaches for extracting novel information from the dMRI signal that could fundamentally transform the capabilities of dMRI.
Although this progress in dMRI is certainly exciting, a common aspect of all fundamental diffusion metrics is that they have no explicit link to tissue properties. The absence of a meaningful construct between these diffusion metrics and specific tissue properties is a significant limitation for the further development of dMRI metrics as biomarkers of neuropathology. By combining diffusion metrics with tissue modeling methods, however, one is able to make estimates for an assortment of tissue characteristics, including the volume fractions, sizes, and compartmental diffusivities for a variety of cellular structures (eg, axons), which may be altered with disease.18–21 With this growing battery of diffusion-based tissue parameters measurable with advanced dMRI, we can expect new biomarkers to emerge for the assessment of neurological disease.
Finally, all of these techniques will benefit from the forthcoming improvements in system hardware and novel acquisition schemes. The most recent generation of clinical MRI scanners offers gradient strengths that are about twice what has previously been available on most commercial systems. These stronger gradients will allow for the design of dMRI sequences with shorter echo times and, consequently, higher signal-to-noise ratio. Multiband echo-planar imaging sequences have also recently become available that allow dMRI data to be acquired with higher acceleration factors resulting in shorter scan times.22
These advances in hardware, acquisition schemes, and approaches to data sampling will increase the feasibility of dMRI techniques in clinical settings, and we should expect dMRI to continue to make important contributions to our understanding of the brain.