Introduction Why Register Images

Images are spatial distributions of information. Accurately relating information from several images requires image registration. Alignment of a PET image with a high-resolution image such as a Magnetic Resonance (MR) image has successfully allowed anatomical or structural context to be inferred from the coarser-resolution PET image. PET to MRI registration was one of the earliest successful examples of image registration to find widespread application. Since then, image registration has become a major area of research in medical imaging, spawning a wide range of applications and a large number of papers in the medical and scientific literature. Recent reviews are provided in Maintz et al. [1] and Hill et al. [2]. Much of this chapter is a summary of information in the latter plus a recent textbook on image registration [3]. Further algorithmic and implementation details are contained in these two sources.

This chapter addresses the software approach to image registration. The first section classifies registration applications and outlines the process of registration. It then discusses some concepts of correspondence inherent in image registration and summarises frequently used transformations. Methods for aligning images based on landmarks or geometric features and recent advances using the statistics of image intensities directly to align images -the so-called "voxel similarity" measures - are described. Some details are given on image preparation, optimization, image sampling, common pitfalls and validation.

Was this article helpful?

0 0

Post a comment