I used MCFLIRT to estimate head motion, then regressed out the motion parameters from individual rs-fMRI data as well as ICA decomposition with removal of noise components, including those appearing from motion. All of my subjects had less than 1 mm total head displacement in any direction, with the majority being even less than 0.5 mm.
I feel like I've done quite enough to take into account motion. But I see that the new way of thinking about head motion is more to do with framewise displacement. Do the FSL tools have a means of incorporating such factors?