---

De : Sean F Walsh <[log in to unmask]>
À : [log in to unmask]
Envoyé le : Mercredi 16 Novembre 2011 17h23
Objet : Re: [FSL] AW: [FSL] AW: [FSL] AW: Is there an FSL-VBM for longitudinal studies yet?

Hi Gwaenelle et al.,

Thanks a lot for the very informative conversation. I have been following the steps suggested in this conversation and have got as far as

6) convolve the subsequent warpfield with the halfway .mat to obtain the final warpfield to apply to each native GM image

Would you mind explaining to me how to do this? I don't want to make any silly mistakes at this stage.

I imagine that this is necessary so that the steps taken in the affine transformation from native to halfway space are taken into account in the Jacobian for the modulation in the following step.


-> Actually, the Jacobian is compensating only for the non-linear component of the registration, so the only Jacobian you need is the one created at the fnirt step. The concatenation between the halfway .mat and the warpfield generated by fnirt is done so that you only interpolate once your data from the native space directly to the standard space. You can do this using something like:

convertwarp -r MNI_152_2mm.nii.gz -m timepoint1_subj1_halfway.mat -w average_subj1_warp.nii.gz -o timepoint1_subj1_to_MNI_warp.nii.gz
 
then apply the new warpfield using applywarp and spline interpolation.


Also, would step 7) modulate and smooth - be as simple as

fslmaths Subject1_GM_TP1_MNI -div Subject1_Jacobian_TP1_to_MNI Subject1_GM_TP1_MNI_mod

-> I'm sorry, the manual is actually confusing and we were supposed to edit it ages ago: you actually need to *multiply* by the Jacobian, so:

fslmaths Subject1_GM_TP1_MNI -mul Subject1_Jacobian_TP1_to_MNI Subject1_GM_TP1_MNI_mod -odt float

fslmaths Subject1_GM_TP1_MNI_mod -s 2 Subject1_GM_TP1_MNI_mod_s2
fslmaths Subject1_GM_TP1_MNI_mod -s 3 Subject1_GM_TP1_MNI_mod_s3
fslmaths Subject1_GM_TP1_MNI_mod -s 4 Subject1_GM_TP1_MNI_mod_s4?

-> First, concatenate in time all your _mod.nii.gz in the right order (corresponding to your design.mat), then smooth it. For the creation of the mask, concatenate all the non-modulated images, take the mean and threshold/binarise at 0.01 depending on your data...

Hope this helps,
Gwenaëlle

--------------------------------------------------------------------