I am trying to examine the diagnositc performance in DARTEL-based VBM. My unmodulated data show larger relevant difference areas and have higher diagnostic accuracy than the modulated. What could explain such result? Why the unmodulated demonstrated much more affected areas than the modulated? Is the following correct according to the definition of how modulation works: "perfectly registered unmodulated data should show no difference between the groups"?Does it mean that unmodulated data have some compound of registration failures? What areas in the brain are more prone to DARTEL-registration failures?
Thank you in advance for your consideration.