Automated CT and MR brain image registration using geometrical feature correlation

Describes an automated, retrospective approach to register CT and MR brain images, using patient related image properties only. Mathematically well founded differential operators in scale space are applied to 2D or 3D image data, resulting in feature images depicting "ridgeness" The CT and MR feature images show similarity which can be used for matching using a fully automatic hierarchical correlation scheme. Results of 2D and 3D matching experiments are presented. >