Dear Tim,
thank you very much for your answer. A few questions, just to make sure
that I got your point.
>1) I'm sure you are not doing this, but just in case: you don't want to
>actually register the eigen* images to the MNI brain. You should register
>either the T2 reference scan, or the mean diffusivity or something
>similar, store the transform file, and apply this transform to whatever
>images you are interested in.
What I actually do is use the 'Low res image -> High res image ->
Reference image' option of FLIRT, where the low res image is the DTI
image, the High res image is a T1 structural one and the reference image
is the MNI template. Is that what you are talking about?
>2) Having said that, you are right to worry about the interpolation.
>Interpolating tensors (or even vector fields) is not at all trivial,
>(e.g. with linear interpolation two vectors in the same orientation but of
>opposite polarity could interpolate to zero!). If you are going to apply a
>transform straight to the eigenvector images, you should do it using
>nearest neighbour interpolation. However, I think what I would do if I
>really wanted my eigenvector images in standard space (which, come to
>think of it, I never have) is apply a flirt transformation matrix to the
>original diffusion weighted data series, and refit the diffusion tensors
>in standard space.
Could you be more specific? Sorry, but I just started using FSL and I am
trying to familiarize myself with the software.
Thanks a ton,
Stamatis
|