ABSTRACT
In Chapter 7, we dealt with all types of spatial transforms applicable to images. Besides visualization and interpolation purposes, these operations provide the core formalism for image registration or image fusion. From the very beginning of this book, it was emphasized that medical imaging modalities do not provide images of the patient’s anatomy and physiology. They do record certain physical properties of tissue. What sounds very philosophic is indeed a very general property of perception. We do only see what we can see – this forms our reality and our image of the world. But modern medical imaging devices enhance the limits of our perception. Less than forty years ago, the introduction of CT replaced invasive
methods like exploratory laparotomy in case of unclear lower abdominal pain; MR gives stunning images of soft tissue where CT fails to achieve good image contrast. US allows for quick imaging not using ionizing radiation, and specialized probes allow even for using US inside the body. PET and SPECT visualize pathologies before these show up as anatomical changes by mapping metabolism and physiology. Still this plethora of information is worthless if it cannot be connected. This is mainly the task of well-trained experts in radiology, but the fact that most modern modalities produce 3D volume image data does not simplify the task of fusing information from different imaging sources. This is the domain of registration algorithms, where a common frame of reference for multiple data sources is established.
