Automatische Volumetrie der Hippokampusformation durch Transformation vordefinierter "Regions-of-Interest"

The present thesis deals with the implementation and analysis of configurable software for the registration of three-dimensional tomographic data and segmenting the Hippocampus. For segmenting Atlas Matching based on linear and affine transformations was used as a simple and time-saving approch. The comparison of different cost functions and the optimization of the configurable parameter of the related specific registration procedure were also part of the work. Image registration is a fundamental task in medical imaging, aiming to align two or more digital images. The data can be acquired, e.g., at distinct times, by different techniques or with distinct perspectives. The image registration makes comparison of images, combined representations of information, or reconstruction of a three-dimensional volume from a series of two-dimensional images possible. The aim of the registration process is the alignment of the objects contained in the images. If the registration is done based on the intensities of the images, then a similarity metric based on the pixels of the images is optimised within the registration process. Therefore we apply the Powell and, respectively, the Simplex optimisation algorithm combined with various metrics and interpolation methods for the registration of MRI images. The measurements can be varied based on the free choice of a function that determines the similarity measure. Functions which have been implemented were the sum of the quadratic intensity difference, the sum of the absolute differences, the mutual information, and the modified ratio image uniformity (MRIU). When comparing the similarity measures, MRIU shows to be, from a global point of view, the best T1-weighted magnetic resonance imaging data. A reduction of the Hippocampus is the part of many diseases in particular for the Alzheimer disease. Due to automatic segmenting and volumetric analysis of the Hippocampusformation in medical imaging a more exact assessment of the reduction of the Hippocampus should be achieved. As Template, a segmented Hippocampus, was provided from the Department of Neurology. However no satisfying result could be obtained by Atlas Matching, because of biological variation and insufficient resolution of the images. But the Hippocampus located from the Atlas Matching can be used as starting point for further optimization algorithms.