Hi Jun,
Look at the rm.fm.VSfent+tlrc dataset, along with follow_ROI_FSvent+tlrc, and see what they show as the ventricles. Maybe they are very small, and the final dataset ends up with no voxels in it. Perhaps FreeSurfer was not able to define them well. You should be sure of why it comes up empty.
There is currently no login for what to do in the proc script if such a mask comes up empty. Since there is nothing to regress, and assuming the mask was made in a reasonable way, but is small enough to disappear, there might be no happy choices going forward (drop subject, more gently erode the mask, substitute your own mask, do not regress from the ventricles).
- rick