Dear Alfonso, 

Thank you for the your answer.
If I understood well from reading the two previous posts you mentioned, it seems that I cannot define ROIs from subject-space but only from fsaverage space. Is that so?

What I have as input is a surface overlay file that is created by morphing a specific atlas to each and every single subject (relying on the subject-specific anatomy).
Should I then morph this overlay to fsaverage to be able to using CONN's analysis pipeline?

Cheers,
Sami