This book presents more than fifty detailed expositions in a half-dozen major areas:

- Enhancement of indistinct images,

- Segmentation into regions with different physiological meaning,

- Quantification of textures, structural features, and diagnostic markers,

- Registration of images from multiple modalities, times, or patients,

- Visualization of complex data in intuitively meaningful displays, and

- Compression and storage of the vast volumes of data in even a single image.



Many chapters apply well-known image processing and image understanding techniques, sometimes with extensions for volumetric data. For example, there's diagnostic information in the angle between two parts of a bony structure, but X-ray images generally foreshorten structures tilted away from the image plane; 3D reconstruction techniques help out here. Segmentation problems take on special urgency when used to isolate lesions or other abnormalities. Techniques often merge in solving any one problem. For example, identifying a brain lesion might first require segmenting the different kinds of tissue using volumetric adaptation of common imaging techniques, warping the standard physiological model to match the patient's unique anatomy, then applying neural networks or other mechanisms to identify the differences of clinical interest. Then, as any mammogram patient knows, non-rigid tissues can shift during imaging so that the picture taken corresponds only roughly to the body structures as they appear normally.



The target reader has rich mathematical background, and some working understanding of image-capture technologies and of diagnosticians' needs. This book aims squarely at the engineer translating the image data into medically useful forms, without losing data of diagnostic importance (and this means you, all you image-compressors out there). By design, this addresses the higher levels of image computation. It deals only occasionally with instrumentation issues, such as image reconstruction. I was interested to learn, however, that the strength of MRI as a tool for chemical analysis is a problem in imaging - the same chemical shifts that represent molecular structure actually mis-represent physiological structure.



This massive book offers a lot to many practitioners in the computations behind contemporary medical imaging. For others, however, the 900-page thickness will be most helpful when a small child sits at a tall table, once she's too big for a booster seat. My immediate needs have to do with image reconstruction from raw data, a topic addressed only tangentially here, so I'm in the second category.



-- wiredweird