Review of "Medical Image Analysis Methods" by Lena Costaridou

The book is organized in 12 chapters. Chapter 1 reviews a particular CAD system for the detection of microcalcifications and masses in mammography. Chapter 2 is similar to Chapter 1 in that it is also concerned with the study of CAD systems for mammographic images. By contrast, Chapter 2 slants towards generality. It spells out the basic elements of a CAD system and then gives a cursory compilation of various processing methods used in the main stages of a typical system. The chapter also discusses the problem of reconstructing the three-dimensional shape of arteries, exploiting both intravascular ultrasound images and biplane angiographies. This chapter could have better served as Chapter 1, as an introduction to CAD systems and related techniques. Chapter 3 reviews another computer-aided system. It focuses on the automated characterization of atherosclerotic carotid plaques from high-resolution ultrasound images. The reviewed computational scheme is based on a neural network and statistical pattern recognition techniques. Chapters 4 to 9 and 11 give either a summary of techniques or a detailed description of methods used for specific tasks usually required in a CAD system. Chapter 4 gives a comprehensive review of classification methods that have made their way into medical image processing. Chapter 5 centers on texture characterization using autoregressive models, while Chapter 6 covers topics in image enhancement using wavelet analysis techniques. Chapters 7 and 8 are concerned with image segmentation. In Chapter 7, the segmentation of magnetic resonance (MR) images via multiscale gradient watershed hierarchies is reviewed. In Chapter 8, the segmentation problem in mammography is examined using Markov random field (MRF) models. Chapter 9 offers an approach to estimate the geometric transformation that puts two medical images into correspondence. And Chapter 11 reviews the problem of how to combine multimodal information, in particular that coming from electroenc