Hi Alfonso,

Thanks for your reply and suggestions. I have gone through all my subjects and checked their normalization, coregistration, WM and CSF masks, and it all looks fine. The WM and CSF are also included as confounds in the denoising tab with 5 dimensions. 

However, at least for some subjects, it does not look like spurious correlations or in subcortical areas. In the attached example, there is large groups of voxels (at threshold .25) in CSF and lesion area. Looking at the GM, WM, and CSF ROIs, it looks like these are all being correctly identified. So I thought the CSF/WM should be masked out of the analysis, or am I misunderstanding something?

Thanks,
Elsa