Hi Paulo,
If you are doing custom preprocessing and want to regress out the voxelwise confounds from PNM, then it should be possible to do this with fsl_glm and the voxelwise confound option, something like:
fsl_glm -i input --vxf=voxelwiseConfound1,voxelwiseConfound2,etc --out_res=output -d blank.mat
should work - the "residuals" output being the original data corrected for the confounds. blank.mat should be a single column of zeros, with a number of rows equal to the number of timepoints ( you can create this with Glm_gui ).
you can use evlist.txt to find the names of the files to be passed to the vxf option.
Note that if input has already been filtered ( e.g. highpass ) the same filtering must be applied to the voxelwise EV's.
Kind regards
Matthew
> Hi everyone
>
> Quick question: how should one go about regressing out the PNM EVs from a 4D dataset? I want to do this as a preprocessing step (rather than including it in FEAT) to be able to use different post-processing strategies. The thing that confuses me a bit is that unlike WM and CSF output from FSLMEANTS, the outputs are actually voxelwise regressors. If I'm correct, this can be done in FEAT creating a voxelwise EV for each .nii file and then looking at the residuals, but I was looking for a quicker, potentially scriptable alternative. Can the fsl_flm command be used for this purpose?
>
> One last question. I have 32 EVS from PNM, to be extracted from 15 minute (300 volumes) resting-state data. Is this recommended, or should I cut down in the EVS outputted from PNM? Any suggestions? I'm already running ICA-AROMA, and will also regress out WM and CSF.
>
> Thanks in advance
> PB
|