White Matter Worries: A Problem for DTI?

By Neuroskeptic | September 1, 2018 12:43 pm

A new preprint called “A systematic bias in DTI findings” could prove worrying for many neuroscientists. In the article, authors Farshid Sepehrband and colleagues of the University of Southern California argue that commonly-used measures of the brain’s white matter integrity may be flawed.

The technique in question, diffusion tensor imaging (DTI), uses an MRI scanner to measure the diffusion of water molecules at different points in the brain. In white matter, water diffusion is constrained by the nerve fibres (myelinated axons) which run in particular directions. So by measuring the tendency to diffuse in some directions more than others (‘fractional anisotropy’), we can infer how many (healthy) axons are in a certain location.

The problem, according to Sepehrband et al., comes from the perivascular space (PVS) – fluid-filled areas surrounding the blood vessels in the brain. Depending on the size of the PVS and the resolution of the DTI scan, water within a PVS could appear to be either directionless or directional, the authors say, and can introduce an insidious bias into studies of white matter.

Given the relatively fast diffusivity of the water expected in PVS, even a small portion of PVS in an imaging voxel can have a substantial effect on the voxel averaged DTI measures

Sepehrband et al. illustrate this concern with simulations and analysis of real DTI data. Their key result is that commonly-used DTI measures are contaminated with PVS bias, and that ‘free water estimation’ (FWE), a popular method intended to prevent contamination of DTI, is not enough to prevent this.

However, all is not lost, because the preprint also introduces a new data analysis method called tissue tensor imaging (TTI) which, the authors say, is able to estimate and correct for the PVS signal.


The bad news is that some DTI findings fail to replicate with TTI. For instance, in the well-known ADNI Alzheimer’s disease dataset, DTI shows that older people who suffer cognitive impairment have lower white-matter integrity in many regions compared to cognitively healthy people. This is typically interpreted as meaning that white matter damage or degeneration has occured in these individuals.

Yet when Sepehrband et al.’s re-analysed the same data using TTI, they found no differences between cognitively impaired and cognitively healthy people, although the impaired group did show increased PVS volume. This suggests that it is actually the PVS, not the white matter, that shows changes in these cognitively impaired cases. (Although the sample size of these analyses was quite small, with 17 cognitively normal and 16 impaired.)

Overall, the authors conclude that they have discovered a problem for diffusion MRI, although they are hopeful that study of the PVS may provide new insights in future:

We show that unincorporated anisotropic free water in perivascular space (PVS) significantly, and systematically, biases DTI measures, casting new light on the biological validity of many previously reported findings. Despite the challenge this poses for interpreting these past results, our findings suggest a new opportunity for incorporating the PVS contribution into diffusion MRI metrics of tissue changes, ultimately strengthening the clinical and scientific value of diffusion MRI.

This article is still a preprint meaning that it hasn’t been peer-reviewed. However, it seems to me that it raises a significant issue and it’ll be interesting to see how other experts in the field react to these claims.

CATEGORIZED UNDER: methods, papers, select, Top Posts
  • Pingback: White Matter Worries: A Problem for DTI? – Numerons()

  • FSE

    This isn’t really a problem with DTI. After all, patients with abnormal DTI still have abnormal DTI.

    The only change is in how we interpret DTI. Abnormalities in DTI were previously thought to indicate rather vague abnormalities in white matter, and now we have a more concrete diagnosis for some cases: expansion of the extracellular space (which implies atrophy/shrinkage of the intracellular compartment).

    Incidentally, anisotropy is not a measure of “non-randomness”, it is a measure of directionality. Imagine particles undergoing Brownian motion inside a large sphere or inside a narrow tube. In both cases the motion of individual particles is unpredictable (i.e. random), but particles in the tube will tend to travel up or down the tube whereas those in the large sphere will not have any directional preference.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Thanks. I have changed the explanation of anisotropy as it was imprecise.

      While it’s true that “The only change is in how we interpret DTI”, can we really be sure that a DTI abnormality is evidence of “atrophy / abnormal shrinkage of the intracellular compartment”? Do we know that every case of PVS enlargement is accomplied by shrinkage or loss of neurons?

      I know that in some sense, assuming the brain has a constant size, the enlargement of one thing must imply the shrinkage of something else. But can we be confident that this shrinkage is always evidence of dysfunction?

      • FSE

        Like you said, if the PVS (in the extracellular extravascular space) is enlarged then something else must be smaller. One possibility is the extracellular intravascular compartment, but it is small to begin with and causes other visible effects (ischemia) when it shrinks. Another possibility is that shrinkage occurs elsewhere in a different extracellular extravascular compartment. I find that implausible. So I think the intracellular compartment is the most biologically plausible: it’s the largest compartment (and therefore has most room to shrink), and it is where brain metabolism takes place after all. Note that “intracellular” does not necessarily mean “neuronal”, most cells in the brain are not neurons.

        Finally, in these studies the inference almost always goes the same way: disease X -> “shrinkage”. You should indeed be skeptical of the reverse inference, because it is not necessarily true that shrinkage is significantly associated with disease X, or any other disease.

    • Farshid

      Thanks Neuroskeptic for writing about our preprint!

      Thanks for opening the discussion FSE. I agree with you to a large extent. DTI is sensitive to tissue changes and the abnormality indeed can be detected.

      DTI sensitivity is not under question here, rather the inference made. In many cases, DTI changes were interpreted as a change in white matter (for example increased mean diffusivity in white matter of AD patients were interpreted as white matter degeneration that leads to more water displacement). Here we showed that the estimated increased diffusivity mainly rises from the PVS diffusion signal and not the white matter diffusion signal. I remain skeptical of any interpretation yet, but I believe it is important to know the source of the signal change.

      Regarding “atrophy / abnormal shrinkage of the intracellular compartment”, I think this is a very interesting topic and differs across diseases. I am not sure if we really know the mechanism or the pathophysiology.



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar