Hi all,
I am attempting to conduct a group comparison between pre- and post-op rsfMRI data sets of the same 7 patients. I have been encountering issues for the past few weeks now, and thus, by this point my most recent work-flow may be a little 'unconventional'. It is as follows:
- brain extract all T1 weighted images (to be used for registration)
- run a single session melodic on one patient’s pre-op fMRI data to get a single session ICA (as regfilt is to be conducted on single session ICA prior to group analysis)
- identify/classify noise components to use for later de-noising step
- run FEAT registration/pre-processing on all 4D_fMRI patient data (did this because the Wiki for regfilt specifies that regfilted data should be pre-processed before subsequent FEAT or group Melodic, and not be pre-processed again by FEAT or Melodic)
- run FSL_regfilt on pre-processed 4D_fmri data (ie. P1_filtered_func_data.nii.gz) from all patients (from previous step), specifying noise components labeled in the previous single session ICA of patient 1
- run Melodic group ICA (with all pre-processing turned off, with the exception of registration) on regfilted_4D_fMRI data
- run dual regression for group comparison between pre- and post-op data (using specifiedGLM)
All steps have appeared to work, with the exception of the dual regression. Is this work flow correct, or am I over complicating things? If someone could specify an appropriate work flow (with brief description of rational behind each step) for such a comparison, I would greatly appreciative.
Look forward to hearing back.
Regards,
Hayden
|