Registration.

Abstract

The goal of image registration is to find a 1-1 point-wise correspondence between two images, a subject image and a target image. Knowing the pointwise correspondence between two brain images allows comparison of structural and functional imaging data such as regions of interest, functional data (e.g., fMRI, EEG, MEG, DTI), and geometric shapes. The image registration process also allows creation of probabilistic anatomical atlases (Mazziotta et al., Neuroimage 2(2):89-101, 1995; Thompson, J Comput Assist Tomogr 21(4):567-581, 1997; Thompson et al., Detecting disease-specific patterns of brain structure using cortical pattern matching and a population-based probabilistic brain atlas. In: IPMI2001. Lecture notes in computer science, pp 488-501, 2001), automatic segmentation by label transfer, modality fusion, morphological analysis (Hua, Neuroimage 43(3):458-469, 2008), and many other applications. Image registration techniques strive to find a one-to-one correspondence between subject and target images to perform this task. This correspondence is defined by a smooth deformation field. This deformation field captures the geometric variations in the two images. In this chapter, we will review various techniques for image registration that are specifically designed for human brain.