Deformable image registration by adaptive Gaussian forces

This paper introduces a novel physics-based approach to elastic image registration. It is based on applying Gaussian-shaped forces at irregularly distributed control points in the image, which is considered to be an infinite elastic continuum. The positions of the control points, the directions and magnitudes of the applied forces as well as their influence areas, and the elastic material properties are optimized to reach a maximum of the similarity measure between the images. The use of the adaptive irregular grid potentially allows to achieve good registration quality by using fewer parameters as compared to regular grids, e.g. B-splines. The feasibility of the proposed approach is tested on clinical images, and open problems and directions for future work are discussed.