Application of Microstructure/Diffusion-Mediated Signals to Study Disease
Matthew Budde1

1Medical College of Wisconsin, United States


This talk will summarize state-of-the-art techniques to probe tissue microstructure and describe ongoing efforts to understand how the biology of nervous system tissue relates to DWI signal and models.


Those interested in using or studying diffusion weighted MRI (DWI) as a biomarker for injury and disease in the brain and spinal cord.


With the information presented, the audience will be better equipped to understand and interpret DWI-based metrics in the context of the tissue microstructure with a focus on neurological disease and injury. The relationship between DWI signals and the underlying pathophysiology will be emphasized. The current status of the field will be summarized and gaps and future directions will be identified with an emphasis on guiding the clinical translation of microstructural imaging.


Diffusion MRI has a remarkable ability to probe the structure of tissues at resolutions far below that of the imaging resolution. It has long history of being a useful to investigate the microscopic structure of the brain and spinal cord, and has a clear and demonstrated utility in clinical diagnosis of cerebral ischemia. However, the promise of using DWI as diagnostic tool in other disorders and injuries has not seen a similar translation from research settings to clinical use. Even well-established methods such as diffusion tensor imaging (DTI) have seen very modest usage in the clinical environment in only a selected situations. This talk will summarize state-of-the-art techniques to probe tissue microstructure and describe ongoing efforts to understand how the biology of nervous system tissue relates to DWI signal and models.


DWI, like many other MR contrasts, lacks specificity to a single pathophysiology. Consequently, the biophysics of DWI changes are not straightforward and are hotly contested. Many mathematical frameworks and biologically-inspired models have been proposed and evaluated. Most rely on several common features and differ in subtle ways. For example, biologically-inspired models incorporate several volume fractions each with characteristic diffusivities. Across almost all models, axons are approximated as non-exchanging cylinders with a negligible intra-axonal perpendicular diffusivity and the extra-neurite fraction is approximated as hindered diffusion within a tortuous environment. Models differ in how they handle changes in diffusivities, intra-voxel fiber dispersions, and the presence (or not) of free or restricting isotropic fractions. Each has various approaches for data fitting and assumptions or constraints. After a brief overview of microstructural DWI models, this talk will primarily highlight efforts to examine their validity in relation to gold-standard measures from histologic imaging or under controlled experimental manipulations.


Distilling the relationship between DWI features and the underlying biology is complex in healthy tissue, and many unknowns further complicate extrapolation in cases of injury or disease. Fortunately, the field already has a long history of insightful studies to understand the mechanisms of DWI contrast. For instance, early studies to identify the source of diffusion anisotropy clearly implicated the role of the axonal membrane rather than structural components such as microtubules or neurofilaments. Many of these experimental paradigms were designed to use ‘strong inference’ to assess competing hypothesis, and the insights gained from these studies have been readily incorporated into DWI models. Similarly-designed studies now and in the future are likely to resolve other unknowns that exist. As an example, the diffusion properties imparted by glial cells (astrocytes and microglia) are largely uncharacterized, and it has been shown that their effects can easily be misconstrued as axons or neurites. Given the high content of glial cells in human brain and substantial responses in disease and injury, a clearer understanding of their effect on the DWI signal will be important.

Importantly, studies that have incorporated gold-standard pathology along with multivariate analysis have also enabled strong and focused conclusions since they have been able to define if microstructural models accurately capture the true tissue structure and just as importantly, where they fail or are ambiguous. A tight coupling between DWI metrics and the biology that universally applies across all situations is unlikely. As an obvious example, DWI performs exceedingly well after acute stroke, but the DWI contrast reverses in a chronic stroke lesion and generally coincides with T2 contrast. Knowledge of the conditions under which microstructural imaging become invalid is critical to ensuring their applicability and interpretability.

While microstructural imaging techniques are valuable as research tools to probe interesting features of the nervous system, their ultimate goal and highest impact will be realized in clinical diagnosis and decision making. In this context, their translational value may be enhanced by:

1) Emphasizing the pathologies known to be relevant to a given disease/injury/entity (axonal injury, demyelination, inflammation, etc).

2) Simultaneously minimizing features from other pathologies (edema) or features (crossing fibers, CSF) that confound those measurements.

Results of studies from our group in the injured spinal cord and others will be described in which this approach has led to refinements in contrast along with more simplistic analysis and interpretation.


Almost all DWI microstructure imaging methods remain as a research tools and have not been adopted to clinical practice. The lack of specificity is often criticized as one reason why translation has need been achieved. However, as guiding principles to move forward, specificity does not need to preclude utility. As in the case of stroke, DWI provided useful patient-specific detection of injury while the mechanisms underlying it were unclear and unresolved. As a reminder, features broadly important for all medical tests are related to the following questions:

Sensitive: Can the metric/technique detect changes relevant to the disease/injury under study?

Unique: Do the metrics provide information that cannot be obtained from other or simpler methods?

Reliable: Are the measures reliable and feasible across time, subjects, and scanners/institutions?

Useful: Is the information meaningful in a clinical context, and will it affect clinical-decision making?

Value: Does the test affect patient outcomes and how many patients are likely to benefit?

These guiding principles can serve as reminders of the benchmarks microstructural imaging will have to achieve to become established diagnostic tests.


DWI microstructural imaging holds promise as the “next-generation” of diagnostic imaging biomarkers. However, achieving translation and adoption will have to clearly demonstrate value and utility. Collective refinement of advanced models and continued research to fill in knowledge gaps will be essential to maximizing the likelihood of success.


No acknowledgement found.


No reference found.
Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)