High-speed high degree-of-freedom spatial normalization for human brain imaging

Regional spatial normalization is an important preliminary step in the analysis of 3-D brain images. The goal is to remove anatomical differences by warping each brain image to match corresponding features in a standard brain atlas. We are developing a very efficient regional spatial normalization algorithm based on octree volume decomposition. The original Octree Spatial Normalization (OSN) algorithm was shown to perform regional spatial normalization in binary brain phantoms in less then 8 minutes with accuracy similar to previously published methods. Several modifications were made in OSN algorithm to optimize it for use with human brain images including automated brain tissue segmentation for tissue classification and feature matching methods with fast cross-correlation. Even with these modifications spatial normalization can still be done in less then 15 minutes for 256 arrays