Deformable models with application to human cerebral cortex reconstruction from magnetic resonance images
Chenyang Xu, Jerry L. Prince · 1999
Constructing a mathematical representation of an object boundary (boundary map-ping) from images is an important problem that is of importance to several active research areas such as image analysis, computer vision, and medical imaging. The focus of this dissertation is to investigate deformable models, a boundary mapping technique that incorporates both image information and prior knowledge about the boundary geometry to extract a meaningful boundary description. A key problem with methods reported in the literature is that they have diÆculties in reliably map-ping boundaries when the models are not initialized near target boundaries or are applied to reconstruct boundaries with concavities. In this research, we make three main contributions to the area of boundary map-ping. First, we developed a method called the gradient vector ow deformable model that is robust to both model initialization and boundary concavities. Second, we developed a generalization of the rst method that allows for improved performance in converging to narrow boundary indentations and greater accuracy in localizing boundaries. Third, we developed a method for reconstructing the central layer of the human cerebral cortex from magnetic resonance images that uses our proposed de-formable model as a core component. Our methods are validated on both simulated images and real magnetic resonance images. This thesis is prepared under the direction of Dr. Jerry L. Prince. ii