Comparison of rigid registration methods in four clinical scenarios

Deformable registration algorithms need in general an appropriate initialization of the volumes that is made by rigid registration algorithms. Very often intensity based rigid registration algorithms are used which provide a good initialization for certain deformable algorithms. However when the algorithm to be used for the deformation is based on a physical approach in which forces are applied to different points of the volume those rigid registration algorithms are not the most appropriate. In this paper we present a comparison between three rigid registration approaches (distance based, curvature based and combination) which detect the areas that are highly deformed and remove them from the calculation of the initial transformation. Its usability is analyzed in four clinical scenarios, namely: livers with large deformations due to breathing, livers before and after tumor ablation, resections and open liver surgery. The results show that the combination approach perfo rms better than the other two in cases when a tumor has been resected.