Registration of medical 4D image sequences an overview of methods

Abstract

The past decade several 4D registration schemes have been developed to align medical 4D image sequences. The complexity of the inverse problem, the many options available to implement all the components of non-linear registration and the variety in characteristics of data sets regarding modality, dimensionality, resolution, size and the process being imaged makes a clear overview of available methods and options necessary. In this paper we discuss distinctive features of 4D medical data, we categorize several aspects of the quality of both an algorithm as the desired outcome. We propose a similarity graph as a visual way of showing between which images the similarity is optimized. And we propose a new scheme to classify 4D registration methods on how they treat the temporal dimension.