Non-linear local registration of functional data
Abstract
Within the scope of three-dimensional brain imaging we propose an inter-individual fusion scheme to register functional activations relatively to anatomical cortical structures, the sulci. This approach is local and non-linear. It relies on a statistical sulci shape model accounting for the inter-individual variability of a population of subjects, and providing deformation modes relatively to a reference shape (a mean sulcus). The deformation field obtained between a given sulcus and the reference sulcus is extended to a neighborhood of the given sulcus by using the thin-plate spline interpolation. It is then applied to the functional activations associated with this sulcus. This approach is compared with other classical matching methods.