The Correction of EPI-induced Geometric Distortions and Their Evaluation

An algorithm for correcting the geometric distortions in the echo planar images and diffusion weighted images is described. In this method the geometric distortion in the echo planar imaging (EPI) data is corrected by non-rigid registration to conventional fast-spin echo (FSE) images. The registration is based on a variant of the Demon´s algorithm [10] with consistency enforced by an improved bijectivity scheme. Quantitative metrics based on the forward similarity and inverse consistency measures, calculated from the closest point distances of the image contours, are used to evaluate the performance of nonlinear registration algorithms. The experimental results demonstrated that the proposed nonlinear registration algorithm is able to correct large distortion in the EPI and diffusion weighted images (DWI) while preserving the topology of brain structure.