Non-linear Registration

PI Name: Jim Gee
Institution: University of Pennsylvania

Abstract:

This Project develops registration methods, compares different deformation approaches and applies these techniques to diverse arrays of diffusion, structural and functional magnetic resonance neuroimaging data. A variety of registration cost functions include completely automated intensity-based or cortically-constrained approaches based on metrics from information theory (e.g., Jensen-Renyi divergence, mutual information) to create population-specific templates and provide guidance in spatial population normalization. This project’s theoretical developments, software extensions and applications aim to provide integrative and multimodal approaches to registration and finding the best subject to subject correspondences. Translational applications of these developments include identification and quantification of long-term effects of prenatal cocaine exposure in adolescents.