[Mristudio-users] Impact of white matter lesionson on accuracy of fiber tracking (FACT)

[susumu mori]

On Mon, May 17, 2010 at 6:53 AM, Yakushev, Igor
<igor.yakushev at uni-mainz.de>wrote:

> Dear Colleagues,
>
>
>
> I wonder if there is any empirical evidence about impact of white matter
> lesions (WML) on accuracy of fiber tracking? The following questions are of
> specific interest:
>
> 1) what is the (at least expected) range of FA-values in a WML? Can
> FA-values within a WML (at least a proportion of voxels) be still within the
> range typical for "normal" WM?
>

> I think it depends. If the lesion is severe like stroke or some MS
lesions, it is filled with fluid. So, FA could be as low as CSF. I think we
can't make a generalized comment.

>
> 2) To what extent can one exclude/minimize impact of WML on performance of
> FACT (in a specific brain region) by simply adjusting the FA-threshold?
>

> Lesion means FA decrease. Then it definitely affect FACT results. FACT is
dependent on FA. If FA gets lower than FACT-FA threshold, you don't get a
tract. Even if you get a tract, FA decrease could lead to higher uncertainty
in eigenvectors. Then tracking through lesion could be more affected by
noise.
> In one sense, the high sensitivity of FACT to lesion is good because it
can detect lesion (although you may not need to do FACT to identify
lesions).
> but in most case, the high sensitivity is not desirable.  That is why we
tested an atlas-based approach (Neuroimage. 2008 Jan 1;39(1):336-47), if you
are interested in MRI-parameter (e.g. FA, MD, diffusivity, T2) along a tract
of your interest
> If you are interested in tract anatomy itself under the existence of
lesions, then the atlas-based approach won't work.
> One possible way to minimize the effect is to use multiple ROIs to
increase the anatomical constraint and lower FA threshold. In this way, FACT
results become less sensitive to FA decrease.

>
> 3) In other words, if one tracts a specific fiber bundle within the brain
> with a pronounced vascular encephalopathy (VE) at a common FA-threshold of
> 0.2, and the resulting track looks fine, i.e. not worse as in a subject
> without any VE, can one conclude that there was just no WML over the course
> of this particular track and tracking is likely to be not biased by VE in
> this subject?
>
> > It is in general difficult to use tractography for quantitative purpose,
but there are two types of information you can get; morphological parameters
and pixel intensity parameters. For morphology, we are talking about the
shape and size of the tracts. You can count the number of pixels occupied by
the streamlines and call it tract volume. The shapes and trajectories are
more difficult to quantify, but maybe you can do group analysis by
normalizing the tract coordinates from the control and patient groups and
compare. For the pixel intensity, you can measure, for example, FA of the
tracts using DtiStudio.
> It is therefore very important to understand if you want to know the
morphology or intensity. In your case, you mentioned "tract looks fine",
meaning you didn't see much morphological differences such as thinner or
more curved. Then when you asked, "can one conclude that there was just no
WML". Then here it seems you are asking more of pixel intensity because WML
may cause FA drops and T2 increase, for example. Then you need to quantify
FA values of the pixels occupied by the tract. You may find that FA of a
portion of the trajectory is actually lower.

4) Let's lower the FA-threshold down to 0.05, repeat the above analysis and
> consider two possible outcomes: a) fiber tracking in the VE+ subject
> produced considerably less fibers than in the VE- subject; b) fiber tracking
> produced the same results in both subject. Can one now conclude that
> tracking of this particular bundle is a) biased b) not biased by VE,
> respectively?
>

> Another inherent complication of tractography-based analysis is the
coupling of the morphology and pixel intensity (FA). Suppose you have a
cluster of pixels with low FA due to WML. Also assume that there is no
change in tract size. If you do fiber tracking with FA>0.2, you may get
smaller tracts in patients with WML simply because of the FA threshold. Then
the result suggests there is a morphological change (the tract is smaller)
in the patient, but the truth is FA drop.This is a common problem of many
MRI-based study. We are observing signals from water and the entire brain
information is contracted to mere tens of megabytes. We can do nice
quantitative analysis of the images but interpretation is not always
straightforward.
> As you mentioned, you can approach the data from multiple angles, like
changing the FA threshold. If the tract is thinner with FA>0.2 but not
thinner with FA>0.05, then it may suggest that the tract is actually there
with the normal size and the FA>0.05 results would yield lower FA values in
the pixels within WML, which is the answer closer to the truth. This would
certainly deepen your understanding of the pathology.
> Tractography is just one way to probe the data and I would suggest you
always combine it with other tools such as manual ROI and pixel-based
SPM-type analysis. Recently we are using atlas-based analysis (see papers by
K. Oishi et al). All these approaches looks at the same data with different
angles. It is very difficult to come to a right conclusion with just one
method.

>
>
>
> Thanks.
>
> Igor
>
>
>
>
>
> ******************************
>
> Igor Yakushev, MD
>
> Department of Psychiatry
>
> University Medical Center Mainz
>
> Mainz, Germany
>
> _______________________________________________
> Mristudio-users mailing list
> Mristudio-users at mristudio.org
> http://lists.mristudio.org/mailman/listinfo/
> Unsubscribe, send a blank email to:
> Mristudio-users-unsubscribe at mristudio.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20100519/a74a5fe0/attachment-0001.html