AUTOMATED IMAGE REGISTRATION - II - INTERSUBJECT VALIDATION OF LINEARAND NONLINEAR MODELS

Purpose: Our goal was to validate linear and nonlinear intersubject im age registration using an automated method (AIR 3.0) based on voxel in tensity. Method: PET and MRI data from 22 normal subjects were registe red to corresponding averaged PET or MRI brain atlases using several s pecific linear and nonlinear spatial transformation models with an aut omated algorithm. Validation was based on anatomically defined landmar ks. Results: Automated registration produced results that were superio r to a manual nine parameter variant of the Talairach registration met hod. Increasing the degrees of freedom in the spatial transformation m odel improved the accuracy of automated intersubject registration. Con clusion: Linear or nonlinear automated intersubject registration based on voxel intensities is computationally practical and produces more a ccurate alignment of homologous landmarks than manual nine parameter T alairach registration. Nonlinear models provide better registration th an linear models but are slower.