Hi Maria,
You can extract the average displacement for each subject within a mask using fslmeants:
fslmeants -i output_from_first_utils.nii.gz -m mask.nii.gz -o displacements.txt
If you want to use the output from randomise, you can threshold that using fslmaths to get a mask. In this case, you will probably want to make sure that your follow-up analysis is orthogonal to your original test.
Cheers,
Eelke
> On 10 Sep 2015, at 10:06, Maria Serra <[log in to unmask]> wrote:
>
> Hi everyone,
>
> thanks Eelke the answers, they have helped a lot.
>
> I have another question regarding shape results.
>
> The command line:
> cluster --in=ConDis_R_Hippocampus_rand_clustere_corrp_fstat1.nii.gz --thresh=0.95
> gives you the extent cluster of the shape differences and its MNI location. However, I would like to know if it's possible to obtain the value of the deformation for every single patient. That way, these values could be entered in a statistical package (e.g. spss) and correlated with clinical variables.
> Many thanks,
> Maria
>
>
> On Wed, Sep 9, 2015 at 5:36 PM, Eelke Visser <[log in to unmask]> wrote:
> Hi Maria,
>
> > I am performing a shape analysis via FIRST. Since I've found shape differences in some brain regions, I was wondering if there is a way to know in which of the 3 possible directions the differences are.
>
> If you used the new style vertex analysis (using randomise), the 4D nifti file that you used contains positive values where the surface for a particular subject is on the outside of the mean shape and negative values where it is on the inside. If you used t-tests, your positive and negative contrasts should correspond to these two cases.
>
> In principle it is also possible to look at the three coordinates of each vertex separately, but this will be very hard to interpret, as it is then no longer obvious whether displacements are perpendicular to the surface or along the surface.
>
> > In addition, as regions studied are functionally subdivided (e.g. Amygdala, Pallidum), is there an atlas to precisely locate those differences?
>
> I am not sure what atlases exist, but you could overlay the stats from randomise on any MNI-based atlas to get an idea of where your differences are in that atlas. Someone else on the list might have more experience here.
>
> > Finally, I would like to know whether or not it is feasible to perform a voxel to voxel comparison of the segmented structures. The rationale of doing this is that with the classic voxel-based analysis, some structures lose a lot of gray matter after segmentation step, while doing it this way, we would avoid such problem.
>
> One could compare the masks produced by FIRST instead of the meshes, but I am not sure what would be gained by doing this; as it is a mesh-based method, all differences will by definition show up on the boundary of the structure.
>
> Hope that helps!
>
> Cheers,
> Eelke
>
>
>
> --
> Maria Serra Blasco
> Institut d'Investigació Biomèdica de l'Hospital de la Santa Creu i Sant Pau, IIB-Sant Pau
> Av. Sant Antoni Ma. Claret, 167
> Tlf: +34 93 5537836
> 08025 Barcelona- Spain
>
> P Abans d'imprimir aquest correu electrònic pensa bé si és necessari fer-ho.
>
>
|