The Computer Vision and Robotics group (VICOROB) of the University
of Girona has been working in mammographic image analysis since
in two main directions: the study and development of
algorithms to detect and characterize micro-calcifications and
clusters of them and, on the other hand, the analysis and
implementation of mammographic image registration techniques.
The use of selected shape-based features was proposed in order to classify clustered micro-calcifications as benign and malignant lesions [111]. The computerized analysis of micro-calcifications was divided into four steps: 1) digitization of mammograms and enhancement of images, 2) detection and localization of suspicious areas using a region growing segmentation algorithm based on Shen proposal [161], 3) extraction of shape-based features for every segmented micro-calcification, and 4) analysis of the features using case-based reasoning techniques. Moreover, and due to its demonstrated relevance for issuing a diagnosis [55], the characterization of micro-calcifications clusters was also studied. It has been observed in a great number of malignant diagnosed mammograms, that the only indicator used to issue a diagnosis was the number of micro-calcifications and their distribution inside every cluster.
Following this work, and in cooperation with the Hospital Universitari Josep Trueta of Girona and the Universitat Ramon Llull of Barcelona, the Computer Vision and Robotics group developed the HRIMAC project [113]. The main goal of this project was to develop a web-based computing tool to allow radiologists to assess the diagnosis of breast cancer from digitized mammograms. Next subsection describes in more detail the architecture of this system.
On the other hand, the group has also investigated mammographic image registration [114,]. Image registration is based on aligning images of the same object (taken at different times or views) in order to extract relevant information. For instance, image registration can be applied to mammographic images to assess internal breast changes. Thus, it is possible to detect the development of possible abnormalities in a particular area of interest.