Hi Josephine,
I'm not entirely sure that I understand, but I'll suggest something based on my best guess.
I assume you have three groups - A, B and controls (C) - and that you have found significant differences in A vs C and B vs C and now want to see how similar A-C is to B-C. That is the basis that I'll answer below. If that's not right then please email again to clarify.
The 4D file is a volumetric representation of the scalar distance, in a standard space (MNI) and this would allow you to calculate the correlation of A-C with B-C by taking the set of non-zero values from each and then doing a standard (non-imaging) correlation of those two sets of values. This would give you a measure of similarity, and should be fairly easy to calculate with matlab, python or even with fslmaths and fslstats.
I hope this helps.
All the best,
Mark
> On 6 Apr 2017, at 16:54, Josephine Heine <[log in to unmask]> wrote:
>
> Dear FSL experts,
>
> I have performed a vertex analysis for two groups (say A & B) following the recommendations in the FIRST User Guide. When compared to healthy controls, each group has shown significant shape deformations. As a next step, I would like to find a measure of how similar these deformations are in group A & B – i.e. I am searching for a way to quantify to which extend the same areas are affected. I am currently unsure whether the following way to proceed is correct and feasible:
>
> (I) From what I understood, the shape analysis uses a fixed amount of vertices. Is there a way to access positions of each of the n vertices (for example in the form of scalar projection values to the mean surface in the 4D file)?
> (II) Since their number is fixed, would it make sense to correlate these n positional values of group A and group B outside of FSL as a measure of similarity?
>
> Any comments or hints would be greatly appreciated! Thank you for your continuous support on this mailing list.
> Kind regards,
> Josephine
|