Biomedical Image Analysis: Model-Based Segmentation and Hybrid Elastic Registration

A central task in biomedical image analysis is the segmentation and quantification of 3D image structures. We have developed new model- based segmentation approaches, which are based on 3D parametric intensity models in conjunction with an accurate, efficient, and robust model fitting scheme. We have successfully applied our segmentation approaches to different biomedical applications such as the quantification of human vessels in 3D MRA and 3D CTA images.
Moreover, the registration of images is an important task in biomedical image analysis. Registration denotes the task of finding an optimal geometric transformation between corresponding images where generally nonrigid or elastic schemes are required. We have developed a new elastic registration approach which is based on the Navier equation of linear elasticity. Depending on the application, our approach allows to use either landmark information, intensity information, or hybrid information, i.e., a combination of landmark and intensity information. The registration approach has been successfully applied, for example, to 3D tomographic images as well as to 2D gel electrophoresis images.