Seed-based resting state analysis in FSL-FEAT

Hi everyone,

I’m looking into ways of doing seed-based resting state functional connectivity with data pre-processed in fMRIPrep. I decided to give FSL a go, and have some questions regarding seed time-series extraction and the set up in FEAT.

  1. I have my functional data normalised to the MNI space in fMRIprep which will be my main input in FEAT. However, I will be extracting the timeseries from my seed in another space (SUIT template) and enter these time-series as an EV to model as shown here. Would there be any possible issues with having the seed time series extracted from a different space?

  2. FEAT has a high-pass filtering option that cannot be disabled (under Data tab). It seems like there is an option to apply these to the EVs as well, which is useful. However, if a confound.txt file is also included for de-noising (under Stats option, confound EVs), would these confounds also get high-pass filtered?

  3. Would the seed time-series included as EVs also get de-noised with the confounds included in the confounds.txt? I am sorry if this is obvious. I am new to FSL and want to make sure I am getting it right. I understand how the seed time-series would be included as an EV in the GLM here but I dont understand how this would work really if these are not de-noised in the first place.

Any help would be appreciated and thank you in advance.

Hi @sev-ince,

Not a direct answer to your question, since FSL isn’t my tool of choice, but I thought it would help to suggest some other resources, such as Nilearn (e.g., Seed-based connectivity on the surface - Nilearn), and CONN also has an fMRIPrep import option.


Hi @Steven,

Thank you for the response!

I am actually not doing a surface analysis. SUIT toolbox is too extract time-series from a brainstem seed.

I thought of CONN toolbox and know how to use it a little but I dont know how to use the pre-derived seed timeseries there. For instance; would these come under the conditions tab?

Thanks again!

Sure, but a lot of the code in the tutorial may still be relevant for you.

You would import your seed in the ROI tab.

Thanks Steven, I will have a look at that and this; Producing single subject maps of seed-to-voxel correlation - Nilearn

I cannot use the ROI option in CONN. As I said, I will get the timeseries from another space, not from the input data. The ROI option in CONN allows you to extract the timeseries from the input data within the ROI you defined as far as I know.