Investigating implicit shape representations for alignment of livers from serial CT examinations

Abstract

In this paper, we examine the use of implicit shape representations for nonrigid registration of serial CT liver examinations. Using ground truth in the form of corresponding landmarks manually labeled by a radiotherapist, we carry out an experiment to determine whether nonrigid registration performs better when applied to the original image data or to images constructed from implicit representations of the liver. We compare a variety of standard regularizers (elastic, diffusion, and curvature), similarity measures (sum of squared differences and mutual information), and weighting factors, using three different implicit shape representations: the Euclidean Distance Transform, the Poisson Transform (based on the expected hitting time of a random walk), and a new transform designed to highlight concavities in the shape.