The purpose of the proposed work is to develop and evaluate methods for non-rigid registration of 3-D diffusion tensor (DT) magnetic resonance (MA) images. The motivation is the spatial normalization of ensembles of DT images of the human brain acquired from different subjects. Spatial normalization enables accurate mapping of imaged properties-in this case, the characteristics of the DT, such as diffusion anisotropy-within the brain in order to assist disease diagnosis. In particular, the aim is to characterize statistically meaningful differences that arise from pathologies of white matter as compared to normals.
Specific aims for this proposal are as follows: 1. To devise a similarity measure for use with DT imagery. A comparative measure of similarity between two diffusion tensors is required to drive the registration algorithm. Preliminary work has revealed some problems with standard tensor comparison measures within this capacity. An alternative measure, which examines a weighted combination of different properties of the full DT, will be evaluated. We hypothesize that the use of the full DT in this way will yield better registrations than similarity based on scalar indices derived from the DT. 2. To develop a method for applying transformations to DT images. Compared to the task of transforming scalar images, the application of spatial transformations (or warps) to DT images is complicated by the fact that diffusion tensors contain orientational information, which must be handled appropriately. Several techniques for applying affine transformations to these images have been developed but need to be validated and extended to more complex transformations. We hypothesize that the most effective method will be one which takes into account the effects of the deformative component of the affine transformation. 3. To exploit fully the orientational information contained in the DT for image registration. Simple adaptations of existing non-rigid registration algorithms do not exploit this important aspect of the DT. Methods for exploiting this information in order to improve the anatomic correspondence in computed registrations will be investigated. We hypothesize that explicit optimization over the orientation of the diffusion tensors is required in order to obtain an accurate registration of DT images. There are an increasing number of clinical applications to which DT-MR imaging is being applied. Examples include chronic stroke assessment, diagnosis of neurodegenerative disorders such as Alzheimer's disease, and multiple sclerosis. This project aims to facilitate such studies by allowing comparisons of DT data from different individuals.
| Yushkevich, Paul; Dubb, Abraham; Xie, Zhiyong et al. (2005) Regional structural characterization of the brain of schizophrenia patients. Acad Radiol 12:1250-61 |
| Dubb, Abraham; Xie, Zhiyong; Gur, Ruben et al. (2005) Characterization of brain plasticity in schizophrenia using template deformation. Acad Radiol 12:3-9 |
| Dubb, Abraham; Gur, Ruben; Avants, Brian et al. (2003) Characterization of sexual dimorphism in the human corpus callosum. Neuroimage 20:512-9 |
| Pettey, David J; Gee, James C (2002) Sexual dimorphism in the corpus callosum: a characterization of local size variations and a classification driven approach to morphometry. Neuroimage 17:1504-11 |
