The long term goals of the proposed work are to aid in the analysis of and to engender novel applications based on structural and functional studies of the human brain in a wide variety of problem settings through the development and deployment of innovative computational approaches that improve the accuracy, objectivity, speed, and ease with which clinical and scientific investigations can be conducted and their results integrated. Specifically, this project aims to investigate methods for brain image matching and tissue classification, and their application to: 1) semi- or fully automated interpretation and quantification of structural (MRI,CT) and functional (PET, fMRI) studies using a deformable atlas that is individualized to the subject; 2) high- dimensional spatial normalization of multiple-subject functional studies to enhance the quality of group analysis; and 3) detailed morphometry and shape characterization of neuroanatomy to facilitate clinical diagnosis and basic scientific research. This proposal, based on extensive evaluations of current matching methods, aims to investigate and implement the following improvements in order to make the methods more powerful and practical for the applications specified above: 1) enhance the accuracy of cortical localization by additionally constraining the matching calculations to align high level features, such as the major tissue interfaces and stable sulcal patterns, and by implementing the non-linear theory of large deformation kinematics to accommodate the highly variable shape of the cortex; 2) develop an MRI partial volume segmentation algorithm that is robust to intensity shading to facilitate regional quantification of structure and function and identification of sulcal contours with which the cortical alignment will be improved; 3) extend probabilistic framework for computational anatomy to address the construction of prior models of brain shape from empirical measurements; and 4) make publicly available a portable and extensible distribution of the image matching software that will allow investigators representing a variety of clinical and scientific interests to rapidly prototype, implement, and maintain customized applications of image matching for their own work.
