DTI exemplifies a commonly observed life cycle of new imaging technology: an initial honeymoon period of unbridled enthusiasm, characterized by frequent publications and presentations, is soon tempered by (valid) questions regarding reproducibility and growing skepticism over real-world applications. Eventually a divide occurs: proponents continue undaunted, using the method and publishing their experience, perhaps acknowledging the limitations, but often not. Detractors become frankly cynical and move on, writing off the method as a flash in the pan when no “killer app” emerges to immediately render the method indispensable. Thankfully, a thoughtful group occupies the middle ground. They recognize the limitations but see them as opportunities for revisions, enhancements, and new directions (pun intended). This is where DTI finds itself today.

The debate over clinical applications of DTI often expands into a larger discussion over the benefits and drawbacks of quantitative imaging. Some argue it is ludicrous to take such an inherently quantitative technique as DTI and subject it to the sort of qualitative descriptions that are the radiologist’s stock in trade. These folks want to see the numbers, and those numbers better be validated (get those scanner-specific, normative data bases ready—and make sure you have one for all ages). Others counter that this is a double standard — after all, T1 and T2 are measureable quantities, yet no one seems to mind that we’ve been providing nothing but qualitative descriptions of T1- and T2-weighted images for decades — why demand more of DTI? We hear a lot about the pitfalls inherent to DTI (watch those crossing fibers!), but you would be hard-pressed to think of an imaging method that doesn’t produce potentially misleading images under certain conditions. We learn to work around these pitfalls until we have something better. The truth is, quite probably, that both sides are at least partially right. Yes, a radiologist armed with experience and good judgment can add value with subjective impressions alone, as we have always done. However, give that radiologist a ruler and he is not as dangerous as the saying goes. On the contrary: if the ruler is made sufficiently accurate, and the measurement is placed in proper clinical (and statistical) context, and if size matters, then the ruler enables added value.

It was the “early days” of DTI when Jellison et al reviewed the method’s principles and described several qualitative imaging patterns observed when a neoplasm alters regional white matter organization. It was hoped that such

descriptions would help predict tumor type and grade, define tumor margins, and optimize biopsies and resections. Many of us who have used DTI in this way have indeed found it to be extremely helpful, and several studies have shown that it improves postoperative outcomes. But it was also hoped that quantitative analyses would eventually validate and objectify these patterns. Unfortunately, quantitative studies of many DTI-based parameters—including mean diffusivity, fractional anisotropy, tensor shape metrics, and axial/radial diffusivities—have been remarkably inconsistent when applied to problems of tissue characterization or classification, especially when addressing individual subjects (versus group comparisons). This failure to date probably reflects the inherent difficulty of the problem, as well as inadequacies of the tensor model and multiple sources of variance in the measured parameters. As we continue in our effort to understand these issues better, we turn towards less restrictive models of diffusion phenomena (eg, diffusion kurtosis or diffusion spectrum imaging), and we account for the unavoidable sources of measurement error as best we can.

Where does this leave us at the moment, now that basic DTI is widely available but the reliability of quantitation remains uncertain and better methods are gradually being disseminated? Can DTI be recommended for routine clinical use, and if so, how should it be used? In 2012 the Clinical Practice Committee of the American Society of Functional Neuroradiology convened a subcommittee to grapple with these questions and draft a set of guidelines for clinical application of DTI. Subcommittee members all agreed that DTI has clinical utility — even when applied qualitatively — but that it should be used judiciously, with a full understanding of its limitations and pitfalls. The issue of quantitation was a bit more contentious and required some compromise. Rather than recommend against the inclusion of quantitative DTI parameters in clinical reports (until issues of reproducibility and validation are further resolved), we urged that such reports at least include an estimate of the statistical variance expected in reported parameters as derived from an appropriate control group. Unfortunately, given that we stopped short of defining what constitutes an “appropriate control group,” this portion of our recommendations, taken out of context, may be misconstrued as an endorsement of dubious practices. For example, I am aware of one instance in which our recommendations were cited to support the use of quantitative DTI to prove the existence of traumatic brain injury in litigation proceedings—this was quite a stretch from the document’s original intent.

With all its limitations and pitfalls, DTI has proven clinical utility in specific applications and it has paved the way for even more promising approaches not far behind. Standardization efforts employing large numbers of subjects across multiple centers are ongoing. In the meantime, controversies notwithstanding, neuroradiologists should allow themselves to be comfortable using DTI today­­ — albeit judiciously and with a watchful eye on related developments amenable to clinical translation as they become available.

References

1. Jellison BJ, Field AS, Medow J, et al. Diffusion tensor imaging of cerebral white matter: a pictorial review of physics, fiber tract anatomy, and tumor imaging patterns. AJNR Am J Neuroradiol 2004;25:356–69
2. Field AS, Filippi C, Kalnin A, et al. American Society of Functional Neuroradiology Guidelines for Clinical Application of Diffusion Tensor Imaging: Recommendations from the DTI Standards and Practice Subcommittee of the ASFNR Clinical Practice Committee. March 2012.

 

Read this article at AJNR.org …