Hi all,
I have a methodological question:
I have run probabilistic tractography and generated tracts for my
subjects. Now I want to extract FA values for group comparison. In some
articles that I have read authors get the mean FA value from under a
binary mask (made from the generated tract by thresholding and
binarizing). I myself would rather do a weighted average, using the
tract as the weights, so that the voxels which constitute the "core" of
the tract and have high relative frequencies contribute to the final
value more than marginal voxels, that have only few fibers passing
through. This would, IMHO, provide a more robust measure of mean FA. I
have implemented the weighted average myself, in C++ (using FSL's
newimage library).
What I would like to hear is your feedback -- is this methodology
(weighted mean instead of simple mean) sound or are there any caveats to
this approach?
Thank you very much.
Miro Drahos
http://brainlens.org
|