Out-of-plane non-linearwarping of a slice into volume

Abstract

Non-linear 3D warping is implemented for registration of a planar (2D) image into an anatomical reference volume space for accurate spatial mapping. The mutual information based automatic warping algorithm is expanded to include the 2D-to-3D mapping of slice images exhibiting localized out-of-plane geometric deformations. Presented in this work is demonstration and evaluation of the control point based 3D thin-plate-spline (TPS) function applied to the correction of local deformations in MR slice images. Automated 3D warping of a slice into an anatomical reference volume space is achieved by optimization of the mutual information cost function calculated from the gray values of the image pair, i.e., the reference volume and a slice. Accuracy in spatial mapping was assessed by using the simulated MR data sets in a standard anatomical coordinate system and locally induced deformations in a slice image with known TPS transformation parameters. The results indicate the sensitivity of the final optimization to DOF and selected control point locations.