Coregistration of MR Imaging and Autoradiography: 2D and 3D Thin-Plate-Spline Image Warping Using Points, Paths and Patches

Boklye Kim, Ph.D.
Jennifer L. Boes, Ph.D.
Charles R. Meyer, Ph.D.
Department of Radiology, University of Michigan Medical Center

Abstract

Images obtained from [14C]-2-deoxy-D-glucose (2DG) autoradiography of a rat brain have been reconstructed into 3D volumes and registered with in vivo MRI volume data of the same animal for the purpose of providing spatially registered, histologic "truth" data. Following implementation of a multivariate optimization algorithm to correct for in-plane deformation, a non-affine registration method employing thin-plate-spline (TPS) warping utilizing user-identified point and path features has been developed and used here to correct for cryomicrotome cutting artifacts. The warping algorithm has been applied to the slice-by-slice (2D) registration of 2DG-autoradiography images for reconstruction of a data volume spatially consistent with the uncut specimen. In 3D the same algorithm was used for registration of the reconstructed 2DG-autoradiography volume with in vivo MRI volume data of the same rat brain where the feature set is augmented to include user-identified homologous surface patches as well as the point and line segment features. The warping technique utilizing thin-plate-spline (TPS) transformations computed from user-identified features consisting of points, paths and patches, provides more accurate registration; specifically the result shows a 26% improvement in registration accuracy compared to the previous application employing affine registration technique.


Dr. Kim received an RSNA Resident Trainee Award for this work. The paper was presented at the 1995 meeting of the RSNA in Chicago, Illinois,November 1995.