Physics-based Parametric Models for Deformable Image Registration

Image registration is a necessary prerequisite for many diagnostic and therapy planning procedures where complementary information from different image sources is required. Non-rigid (deformable) registration methods are of particular interest, since deformations of anatomical structures often have a very complex structure and cannot be modeled using simple transformation classes, for example rigid or affine transformations. One well-established classification scheme of the deformable registration approaches is their division into non-parametric and parametric methods. The talk is devoted to physics-based parametric models for deformable image registration based on special analytical solutions of the equations of linear elasticity. These models can be used in conjunction with point-based as well as voxel-based similarity measures or a combination of those. We present the underlying theory, derivation of the models, and implementation issues, as well as show practical application examples with 3D clinical data.