2D/3D non-rigid registration of volumetric CT lung data using thin-plate spline

In the medical image analysis for the human organs, non-rigid registration is necessary to properly register the local deformation. Especially for the dynamic organs like a heart and lungs, it is essential to process the registration method in three-dimensional space to track their through-plane motion. In this paper, we suggest the 2D/3D non-rigid registration method based on the corresponding anatomical landmarks using thin-plate spline (TPS). By extending two-dimensional TPS, we developed 3D TPS and computed its bending energy function to register the local deformations of the volumetric time-varying lung CT data. Our experimental results show that our method is suitable for handling volumetric deformation with a small number of anatomical landmarks.