I am a novice and a PhD student in the rs-fMRI field and would like to learn more about dual regression. I have performed dual regression on several different datasets, and realize that significant results (tfce_corrp outputs ) can be outside of their respective ICA components. Is that something to be concerned about? And how should we interpret these results? Have I done something wrong?

I have attached a sample screenshot of what I mean.

Here is a brief description of my methods. First I preprocess with fMRIprep, and then denoise with XCP-D, and feed the denoised image into melodic. Depending on the subject number, components may vary (the example shown was 15 components on 27 subjects), and datasets temporally concatenated. Voxel-wise comparisons were carried out with 10000 permutations.

A related question, if my design is repeated measures between two groups, should I still temporally concatenate all the datasets?

Thank you!

Regards,

Vae