Effects of Nonrigid Registrations on DBM Analysis using SSD Model

Zhaoying Han^1,2, Xue Yang¹, Bennett a Landman^1,2, John C. Gore², Benoit M. Dawant¹

Deformation Based Morphometry (DBM) aims to detect morphological differences between groups based on statistical analysis of deformation fields generated by non-rigid registrations. Few studies have compared the effects of registration algorithms on DBM. We simulated two groups of normal brain images and deformation fields using the Statistical Simulation of Deformations (SSD) model, introduced known growths in one group, and evaluated the effect of five non-rigid registration methods (ABA, IRTK, FSL, ART, and SPM normalization) on DBM analysis of the groups without and with growths. The ground truth Jacobian map from the known growths was used as the basis for comparison.