I’m having issues with FSL/ICA-AROMA removing all signal in a paradigm known to have robust activation in at least the motor regions. I would like to run my preprocessing steps by the community at large to see whether it’s the processing method or just noisy data that is causing the problems.
My data were captured using opposite encoding directions for each run and I created fieldmaps using method 2 of the instructions I found here https://lcni.uoregon.edu/kb-articles/kb-0003
Following the ICA-AROMA manual, I then used FSL for preprocessing. I deleted the first two volumes, Applied motion correction (MCFLIRT), B0 unwarping, slice time correction using a slice timing file, bet, spatial smoothing (6mm), and registration to the structural (Normal search, BBR) and standard space (12 DOF linear, nonlinear warp resolution of 6mm). I did not apply highpass filtering per instructions.
I then ran ICA-AROMA, simply pointing it towards the feat directory. It ran with no problems. I brought the nonaggressive image back into Feat to complete highpass filtering and registration to T1w and MNI. Originally, that’s all I applied after ICA-AROMA. Like I said, this method resulted in no activation anywhere at all after voxelwise or clusterwise correction was applied.
Next, after reading that the transformations don’t actually get applied to the images until first levels (can someone verify this?), I decided to re-run the second preprocessing step using B0 unwarping, slice timing, highpass filtering, motion correction, registration to T1w and MNI, then ran stats using FILM pre-whitening. Again, nothing came out.
I’ve ran this processing pipeline on scans in both directions with all iterations of warp directions (y, -y, PA, AP), but the images still don’t show activation after voxelwise thresholding.
Like I said, I’m not convinced that my steps are correct, although I haven’t been able to pinpoint a particular issue. Can anyone give a run-down of a method using Feat and ICA-AROMA that should work so that I know that’s not the issue?