FSL ROI analysis

Software Support
1

Hello, I’m new to FSL. This is most likely a very simple question but I’m not entirely sure what to do.

My goal is to extract average activation within a functional roi (obtained in a localiser scan for each subject). Each subject completed 6 separate runs.
From what I’ve read online, the usual way is to first, analyse individual runs in first-level analysis, then combine runs within subjects in second-level, and do group level analysis in third-level.

I’ve thought of three ways to get average activation within roi, but I am not sure which one is the best/correct.

  1. Since the roi is in subject fmri space, I was thinking about turning the cope of each subject (obtained in second level) from standard space to subject fmri space, and use fslmeants to get the activation. The cope is obtained from 6 first-level runs so randomly picking a standard2examplefunc.mat for transformation feels rather wrong… (please let me know if I am wrong)

  2. I could get the activation of each separate run, get 6 numbers and average them. Am I correct in thinking that the ‘combination’ of first-level runs in second-level isn’t simply averaging things so this method is grievously wrong?

  3. Would it be alright if I simply change the roi in subject fmri space to standard space, and slam the standard space roi of each subject to their corresponding cope?

Thank you!

2

I’m not an expert in ROI analysis, but I’ll try to address your questions.

1- If you are using FSL, you should use Featquery to pull signal from your ROIs. If registration has been done for your subjects, Featquery automatically uses the matrices to move between spaces ,so you don’t have to. (If you have used fmriprep to preprocess your data, your L1 folders are probably already in MNI space, so there shouldn’t be an issue.)

2- I am guessing that, you are indeed just averaging things (though maybe some corrections are done). This can be empirically verified by running featquery on L1s separately, and then on L2, to see if you get the same results by averaging.

3- Sounds okay too.

Good luck!

1 Like
3

Thanks Mkassaie! For method 2, there is a slight difference between the results. Perhaps I will stick with method 1 or 3.