Medical Image Retrieval Based on Plaque Appearance

Images provide a powerful means to represent data, and many applications have as fundamental components the acquisition, processing and storing of huge amount of images. A typical example of such application is medical imaging. In hospitals working with medical images, large amounts of image data are received daily for processing, analysis and archiving. This arises the necessity of constructing database management systems able to organize, analyze and retrieve this type of data. The first image retrieval systems were based on information retrieval methodologies applied to textual annotations. As it has been seen in the last decade, describing images by textual annotations is not satisfactory for three main reasons: i) there is too much information in an image, so that small sets of words cannot describe it well; ii) labelling an image by words is a subjective task, as different individuals can apply different labels (words) depending on what they consider more relevant in the current image; iii) a complete manual labelling in a large collection of images is a very tedious task. These reasons have led to the use of image content for processing and organizing the database, constituting the so-called content-based image retrieval systems. Such content-based image retrieval (CBIR) systems are necessary for a wide range of applications dealing with vast volumes of images. In the medical image field, retrieval by content is necessary for performing guided diagnosis and therapy. Having medical images archived along with their descriptions, retrieval by content allows the physician to present a query image representing