Dieter Hahn

A Practical Salient Region Based Hierarchical Method: for Aligning 3D Medical Images Dieter Hahn

Recently, a novel hybrid image alignment method for 2D images was presented, which is based on configural matching of automatically extracted, scale-invariant salient region features. The approach combines advantageous aspects of both feature and intensity based methods. This work proposes an extension of the algorithm to 3D. Its applicability is discussed on both mono- and multi-modality medical 3D images for change detection, tumor localization or time based intra person studies. We describe how results of the rigid body alignment approach can be incorporated into a subsequent non-linear registration. Besides the practical work, the work presents an overview of current research in the field of rigid and non-rigid hybrid image alignment methods and existing patents. The following problems are discussed: an overview of hybrid image alignment methods, the implementation of the salient region based matching method for 3D images, a statistical optimization framework that is suitable for the salient region based 3D image matching, and possibilities for an integration of a non-linear alignment method.