Outline of Deformable Image Registration for Clinical Use

Abstract

DIR (Deformable image registration) is a registration technique which generates a displacement vector field (DVF) to move each pixel of a moving image to the corresponding pixel position of the target image, and deforms the moving image to match the target image. This article focuses the algorithm and fundamental matters of DIR by introducing intensity-based, contour-based algorithms, and hybrid algorithm. The intensity-based algorithm calculates the DVF by optimizing the index that measures the degree of coincidence of the intensity of two images. On the other hand, the contour-based algorithm calculates DVF by applying a physical model that moves the contour points on the moving image to the corresponding contour points of the target image. Recently, hybrid methods using both image intensity information and contour information can also be used. Factors influencing accuracy of DIR include the algorithm, parameters of the algorithm, the combination of image modality, and user's skill. In order to judge whether the result of DIR is clinically acceptable or not, it is necessary to evaluate the accuracy of DIR comprehensively by visual and quantitative evaluation for each case.