Nonrigid image registration is a natural extension to rigid registration, allowing also deformations, in order to achieve a good match. This is necessary in all instances of patient motion within an image, such as respiratory or heart motion. It is also necessary if datasets from different patients should be registered, for instance in atlas registration applications.
The algorithms developed in the pattern recognition lab are based on a nonrigid, nonparametric formulation and are used in applications concerning general motion compensation, PET/SPECT difference imaging and atlas registration for PET-MR attenuation correction.
The animation on the right shows a nonrigid registration process, on an MR T1-T2 brain scan from different patients. Top left: deforming moving image; top right: contour overlay with reference image; bottom: x and y component of the deformation field.