Non-Rigid Registration of Vessel Structures in IVUS Images Jaume Amores and Petia Radeva ⋆ Computer Vision Center, Dept. Inform `atica, UAB Bellaterra, Spain jaume@cvc.uab.es Abstract. We present a registration and retrieval algorithm of medical images. Our algorithm is oriented in a general fashion towards gray level medical images of non-rigid bodies such as coronary vessels, where ob- ject shape information provide poor information. We use rich descriptors based on both local and global (contextual) information, and at the same time we use a cooperative-iterative strategy in order to get a good set of correspondences as well as a good final transformation. We focus on a novel application of registration of medical images: registration of IVUS, a promising technique of analyzing the coronary vessels. 1 Introduction There is a wide range of applications of medical image registration and we refer to books such as [7] for detailed information. We apply registration to IntraVas- cular UltraSound images (IVUS), a powerful imaging modality for analysis and diagnosis of coronary vessels ([1]). In concrete we present a registration proce- dure to be used as a first step in a more general retrieval framework. The IVUS technique produces images with quite particularities and noise, difficult to ana- lyze. Thus, creating a retrieval system of IVUS images is of high clinical interest for diagnosis purposes. Although there is a huge number of works in the area of Registration and Retrieval of Medical Images [2, 7], matching of IVUS images and retrieving cases from an IVUS images database is a new problem to be solved. On the other hand, many works on medical image registration are focused on rigid parts that justifies rigid registration. Medical images of non-rigid bodies such as coronary vessels in IVUS present features quite different as they do not have any characteristic spatial configuration forced by the bony structure. We perform elastic matching with a variational approach for the transformation, given the high variability inter and intra subject of our medical images. Registration consists on finding structures analog in a pair of images and compute a transformation that align them. We will follow point mapping as a general procedure of registration [5, 1]. ⋆ This work is supported by Ministerio de Ciencia y Tecnologia of Spain, grant TIC2000-1635-C04-04