Nonlinear image registration is an initial step for a large number of medical image analysis applications. Optical
flow based intensity registration is often used for dealing with intra-modality applications involving motion
differences. In this work we present an energy functional which uses a novel, second-order regularization prior
of the displacement field. Compared to other methods our scheme is robust to non-Gaussian noise and does not
penalize locally affine deformation fields in homogeneous areas. We propose an efficient and stable numerical
scheme to find the minimizer of the presented energy. We implemented our algorithm using modern consumer
graphics processing units and thereby increased the execution performance dramatically. We further show
experimental evaluations on clinical CT thorax data sets at different breathing states and on dynamic 4D CT
cardiac data sets.
Studying complex thorax breating motion is an important research topic for accurate fusion of functional and anatomical data, radiotherapy planning or reduction of breathing motion artifacts.
We investigate segmented CT lung, airway and diaphragm surfaces at several different breathing states between Functional Residual and Total Lung Capacity. In general, it is hard to robustly derive corresponding shape features like curvature maxima from lung and diaphragm surfaces since diaphragm and rib cage muscles tend to deform the elastic lung tissue such that e.g. ridges might disappear.
A novel registration method based on the shape context approach for shape matching is presented where we extend shape context to 3D surfaces. The shape context approach was reported as a promising method for matching 2D shapes without relying on extracted shape features. We use the point correspondences for a non-rigid thin-plate-spline registration to get deformation fields that describe the movement of lung and diaphragm. Our validation consists of experiments on phantom and real sheep thorax data sets. Phantom experiments make use of shapes that are manipulated with known transformations that simulate breathing behaviour. Real thorax data experiments use a data set showing lungs and diaphragm at 5 distinct breathing states, where we compare subsets of the data sets and qualitatively and quantitatively asses the registration performance by using manually identified corresponding landmarks.