Zilken, H. R. (Corresponding author)

1999
Forschungszentrum, Zentralbibliothek Jülich

Report No.: Juel-3649

Abstract: Three-dimensional anatomical data are very important for medical diagnostics. It is often possible to characterize important properties of human organs represented in these data by a description of their surface, Examples of such properties are the location of the organs in the human body, location relative to other organs, size, shape and volume, and also the curvature of the surface. For a proper determination of these values an exact representation of the surface must be found. In this thesis, the concepts of active surfaces are used to calculate such a representation. By the application of user-given forces, a deformable template is fitted to a surface in an iterative process. This surface is implicitly given by the voxel-data, The process of deformation is guided by inner and outer energy terms. The inner energy is a measure of the stretching and the curvature of the template and is needed to minimize the distortion of the template in the deformation process. The outer energy is calculated from the measured voxel-data and ensures the fitting of the template to the data. Different possibilities of defining the energies and the derived forces are discussed. A discrete method as well as a method based on the principles of continuum mechanics are presented. The latter is discretized by the finte-element method. Furthermore a concept for the parallelization of the algorithm on a heterogeneous computersystem, coupled by a network, is presented. The algorithm has been tested and applied to the calculation of the mapping from the surface of the human brain onto the surface of a sphere. This transformation is non-linear and uniquely maps every point of the brain-surface to a point of a sphere. By means of this transformation it is possible to visualize several properties of the human brain-surface on a clearly arranged general map

Keyword(s): computational neurobiology ; brain imaging ; surface ; finite element method ; parallel algorithm

1. Institut für Neurowissenschaften und Biophysik - Medizin (INB-3)