[Mristudio-users] Fa values

[susumu mori]

Hi Jacqui,

This is a typical question of a coupling between morphology and pixel
intensity (contrast) measurements. The notion of "morphology" is based
on a cluster of pixel. There is no concept of morphology if you pick
one pixel. If you define a group of pixels, then you ask morphological
questions such as shapes and volumes. For pixel intensity, you can
measure it for single pixel as well as a cluster of pixels.

For example, you extract one axial slice at the pons where you can
identify the CST clearly. When somebody ask you to measure FA, you
need to define a boundary. If you define only the core of the CST, the
FA goes up and the size goes down. If your definition includes the
boarder of the CST, the FA goes down and the size goes up. The same
image, two different results; the former method may tell you "this
patient has smaller CST but FA is preserved" and the latter method
tells you "this patient has lower FA but the size was preserved".

So, the question is which is correct? Unfortunately, MRI/DTI can't
answer the question. You are asking a question such as "what is the
status of the CST", which calls for cellular-level information, while
we have merely 20MB of data for the entire brain based on water
diffusion property. Information is already degenerated too much to
answer your specific questions about neuroanatomy.

On top of it, even if we have a histology slice of the patient, it is
very difficult to define the CST because many brain structures do not
have a sharp boundary. The axons comprises the CST is very dense at
the core but toward the edge there are many pontine crossing fibers
interdigitating the CST axons. Even with histology, we can't clearly
define the boarder of the CST, while any image-based quantification
methods require us to define a boarder. So our effort for image-based
quantification is in some sense an atrifactual effort.

Then, the question is, why we do image-based quantification if it
gives results that can only be remotely connected to actual cellular
entities. Well, using MRI/DTI, you can systematically contract the
astronomical entity to a mere 10-20MB matrix, with which you can
readily quantify values and compare subjects and groups. In addition,
it takes only 5 min and non-invasive. While we are enjoying all these,
we really can't complain that it doesn't give us a straight answer
about what is the real cellular events happening in a patient.

Now, going back to the two different results from the same image, all
you can do is to apply the same criteria when you define a structure.
When we draw ROIs to define a boundary, we should try to be
consistent. Using FA threshold is one good way to remove
inter/intra-rater variability. As long as we apply consistent and
reproducible quantification approach, the data are good. However, if
we make mistake, that is when we interpret the data. Even if the
results scream that "No size change, FA goes down!", maybe what is
happening is the reduced size without FA change. Or histologically,
the former would suggest myelination loss or axonal loss, but in
reality it could be the reduced number of axons. So, we just have to
be careful when we interpret the data. Again, in MRI/DTI, the
anatomical data is astronomically degenerated. It is not always
possible to conclude what is happening in cellular levels.

So, to answer your question, yes, measure FA of the CST which is
defined by FA threshold is a tricky approach. Because if there is FA
loss, you may end up in a smaller CST with unchanged FA value simply
because you extracted only high FA portion of the CST. In this case,
you detected FA change by changes in size; a good example of
morphology-contrast coupling. If you find a consistent differences
between the control and patient groups, it's a good start. You just
don't wan to immediately conclude like "therefore, the CST atrophy" or
"therefore FA changed".

>From purely methodological point of view, you can also warp our
probabilistic CST location and measure FA values along the CST. In
this way, you don't perform tractography in each subject. You transfer
the population-based CST definition to each subject. In this way, you
can "freeze" the morphological information and perform pure contrast
detection. However, the downside is, you invite a nuisance factor,
called registration accuracy.

There is no perfect quantification method for image analysis. We just
have to know pros&cons and it is advisable to use multiple approaches.

Susumu



On Fri, Apr 8, 2011 at 12:56 PM, Jacquie Hodge
<Jacquie.Hodge at albertahealthservices.ca> wrote:
> Good morning everyone,
>
> I am using DTI studio for my project where I am interested in comparing the
> left and right CST's in children who have suffereed injury in that area and
> after reading many studies on DTI decided to compare average FA values of
> the tracts as one of my variables. However, I have come across this questino
> from one of my supervisors and cannot really answer it and am hoping for
> your help/expertise.
> Here is goes:
> Is the FA value still a viable measure to look at and use to make a
> correlation if you are using FA values to pull the white matter tract? It
> seems like if the tract is made from FA values then you should not be able
> to use the FA as an outcome variable. This question has come up as I
> presented my research idea and I'm not sure how to answer it. Most studies
> that use DTI use FA thresholds to pull the tracts and then also the FA
> values (eg. mean FA) to make comparisons between the tracts.
>
> I really hope that my question makes sense as I have found it very hard to
> word properly.
>
> Thank you very much,
> Jacquie
>
> Jacquie Hodge MSc candidate in Neuroscience, BSc
> Neuroscience, University of Calgary
> Research Assistant, Calgary Pediatric Stroke Program
> Alberta Children's Hospital
> 2888 Shaganappi Trail NW, Calgary, AB, Canada T3B 6A8
> Phone: 403-955-7733 Fax: 403-955-2922
> jacquie.hodge at albertahealthservices.ca
> Program website:
> http://www.calgaryhealthregion.ca/programs/stroke/pediatric.htm
> ________________________________
> This message and any attached documents are only for the use of the intended
> recipient(s), are confidential and may contain privileged information. Any
> unauthorized review, use, retransmission, or other disclosure is strictly
> prohibited. If you have received this message in error, please notify the
> sender immediately, and then delete the original message. Thank you.
>
> _______________________________________________
> 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
>
>