Dear experts,

I am using a large (n=2000+) resting-state fMRI dataset and have a few questions re: how to analyze such a large dataset.

Some more details: for each participant, I have up to 4 runs of rs-fMRI data that were acquired within one session. The fMRI sequence was broken up to minimize motion and fatigue during the scan. I do not mean to compare runs, but am simply interested in an overall association between a variable (exposure, so equal for every run) and ROI-whole brain connectivity.

I understand it is not recommended to simply merge preprocessed data, so for each run, I have run a lower level FEAT analysis to compute ROI-whole brain connectivity maps. I was planning on running a second-level FEAT analysis to compute COPE’s for each participant, which I would then, in a third level analysis relate to my exposure variable, but the GLM GUI can’t handle the size of my dataset. I have been able to manually create my design.mat and design.con files, but I am not sure how to feed these into my design.fsf file. Is there a way to do this? Otherwise, can a group level GLM be performed from command line? (I have been looking into fsl_glm but am not sure if this can be used for higher-level analysis, e.g. how do I specify if I would like to run a fixed effect or FLAME analysis?)

As an alternative I have been trying to think of ways to analyze my data using randomise or PALM. With regards to randomise, I read on the GLM page that I could use fslmerge to merge all lower level statistical maps for one person to create average per person statistical maps. I also read that I can feed a similar design matrix to randomise as I would to FEAT (meaning all separate runs as inputs, with an EV per participant) but adding my covariate to the designmatrix. Would this work in my case, since my covariate is the same for each run? Do I also understand correctly that I should adjust the design.grp file to represent a “participant count”, so each dataset for participant 1 gets a 1, each dataset for participant to gets a 2, etc. Which of the two strategies would be recommended? Would this also work for PALM?

Thank you very much for your help!

Kind regards,
Sandra



########################################################################

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1