Fast Nonrigid Image Registration using Salient Region Feature and Free Form Deformations

2 Introduction Image registration aims to spatially align one image to another and establish point-to-point correspondences between the images. It is an active and important topic of research in medical image analysis because of its numerous applications including change detection in longitudinal studies, multi-modal image fusion, motion correction in time series, image-guided intervention, anatomical atlas generation and group analysis. Registration happens at both global and local levels. Global registration methods recover parameters of a global transformation model, such as rigid and affine, to account for global differences such as translation, rotation, and scaling. In many cases of medical image registration, local registration that utilizes an elastic or non-rigid deformation model is also required to cope with local differences between the images. Such local differences are due to, for examples, scanner-induced deformations, movement and/or respiratory motion of the patient, surgical intervention, or different anatomy (e.g. in atlas generation). This chapter will review the fundamentals of medical image registration, and describe a novel efficient non-rigid image registration method. The method is based on finding salient region-feature correspondences between images and solving for parameters of a Free Form Deformations (FFD) model in closed form.