Dear FSL experts,
I would like to use voxelwise EVs in my Resting State Network Analysis. I did the preprocessing of all EPI images including FLIRT and FNIRT (as it is implemented in Melodic). Afterwards I ran the command feat_gm_prepare 4D-GM-output ICA1.ica ICA2.ica ICAx.ica…..
1. After the gm_prepare command is finished it is recommended to apply additional smoothing to the 4D-GM-output file. Is it right to just apply fslmaths 4D-GM-output.nii.gz –kernel gauss 2.1233226 –fmean smoothed_4D-GM-output if I have a smoothing of 5mm in my functional images (together with resampling the voxel resolution to match it to the functional images)?
2. My resulting 4D-GM-output image looks like the ones attached to this message. Just to make sure I got the principle behind the 4D-GM-output file. The image shows the variance of gray matter within my sample of subjects (meaning that it is not like a gray matter probability map). So, the higher the values (within the image, the light blue parts), the higher the variance between subjects in terms of GM. Is that right?
3. Further the images show that my 4D-GM-output.nii extends the boarders of the MNI152 template. Does this make sense, or did I do something wrong in my Analysis?
4. I would like to implement 4D-GM-output as a voxelwise EV into Randomise as a confounding regressor. And I do not use the GUI. I have one group of subjects with some behavioral EVs and the 4D-GM-output.
Design.mat design.con
Group EV1 EV2 EV3
1 12 1 1 0 1 0 0
1 13 2 1 0 0 1 0
1 14 1 1
1 15 2 1
Command:
randomise -i melodic_IC.nii.gz -o output -d design.mat (see above, with EV3 being the column for the voxelwise regressor) -t design.con --vxl=3 --vxf=4D-GM-output -n 5000 –T
Is it right to include column with only “1’s” and to use --vxl=3 to replace EV3 with the voxelwise EV?
Thank you very much in advance
All the best
Christiane
|