Technologies for high resolution structural imaging of the brain provide neuroscientists with tremendous opportunities for measuring the brain quantitatively and accurately. There is a need, however, for reliable true inter-subject comparison and accurate automated region-of-interest definition. Limitations in these areas have impeded progress in neuroscience in the understanding of normal and abnormal brain structure and in the characterization and treatment of an array of brain disorders, including posttraumatic stress disorder (PYSD) and autism, two conditions having subtle yet distinctive structural abnormalities. The objective of this work is the development of an automated computer system for the analysis of the brain from structural images from magnetic resonance (MR).
The aim i s to develop and validate methods and provide a tool to allow for accurate, detailed, inter-subject comparison of structure in the brain. Such methods would also be directly applicable to associated functional image. This work requires the refinement and development of segmentation and non-rigid registration methodologies. The limitations of current approaches to registration in terms of lack of detail and inaccuracy can be overcome by incorporating prior statistical information of shape for boundary finding and integrating this image-derived information using physical models for deformation. Characterizing neuroanatomic variation in a detailed probabilistic manner is crucial for constraining the necessary transformation and making it consistent and reliable. The resulting representation allows detailed morphologic description in terms of geometry and deformation. This methodology would also be applicable to comparison within an individual where there is structural change over time. This system will be validated with synthetic, phantom and real images from separately funded studies of PTSD and autism. The results are ideally suited for statistical analysis of structure. This system will allow the accurate and routine analysis and comparison of brain imaging studies to compute regional structural measures for improved understanding of the normal and abnormal brain and in the treatment of patients with brain disorders.
| Duncan, James S; Papademetris, Xenophon; Yang, Jing et al. (2004) Geometric strategies for neuroanatomic analysis from MRI. Neuroimage 23 Suppl 1:S34-45 |
| Wang, Yongmei Michelle; Schultz, Robert T; Constable, R Todd et al. (2003) Nonlinear estimation and modeling of fMRI data using spatio-temporal support vector regression. Inf Process Med Imaging 18:647-59 |
