Dear FSL experts,
I wanted to process some resting state that were already preprocessed in AFNI (originally for seed-based analysis) using Melodic for ICA. In terms of preprocessing I did the following:
1) Despike
2) Physio correction of cardiac/respiration
3) Slice timing (TR =3 sec so did not feel fast enough to warrant going after motion correction but I could be wrong)
4) Motion Correction and also calculate motion parameters/derivatives of motion parameters
- Also calculated are timepoints to discard due to high motion (> 0.5mm) or extreme outlier (not taken out but just a file created of which to take out later)
5) FUGUE
6) Anaticor - Regress out motion parameters, WM, CSF, Saggital Sinus
Here is where I have a few questions to see how seed-based and ICA preprocessing deviates from each other:
7) Question at this stage: Do you recommend global signal regression for ICA or not?
8 )Question at this stage: Do you recommend bandpass filtering or not (0.008 - 0.1 Hz)? From what I read it may not be recommended for ICA...
9) Question at this stage: Do you recommend smoothing of resting state data? If so,do you smooth within a GM/WM/CSF mask separately (I have from freesurfer aseg file) or do you smooth across the whole volume?
10) Question: Typically I would run a command called 3dDeconvolve on the clean time series (after stage 9) with my seed region. I would introduce my censor file to discard timepoints (bullet on stage 4) in both my seed and cleaned timeseries . Would this type of motion scrubbing still be valid with ICA analysis? If so, at what stage would you implement it? Would I need to replace the timepoint with another value for ICA (as I do not have to do this with seed-based), and if so do you have a tool which can do this if I supply a text file of 1's and 0's of which timepoints to keep.
Aside from the questions above, are there any changes you would make for the pipeline (ideally if starting from scratch for a new study/FSL from the beginning so wanted to see if the general steps are the approach you would take)
Thanks,
Ajay
|