International Journal of Neural Systems Special Issue on Issue’s Topic c  World Scientific Publishing Company PATTERNS OUT OF CASES USING KOHONEN MAPS IN BREAST CANCER DIAGNOSIS A. Fornells 1 , J.M. Martorell 1 , E. Golobardes 1 , J.M. Garrell 1 and X. Vilas´ ıs 2 1 Grup de Recerca en Sistemes Intelligents, Enginyeria i Arquitectura La Salle, Universitat Ramon Llull 2 LIFAELS, Enginyeria i Arquitectura La Salle, Universitat Ramon Llull Quatre Camins 2, 08022 Barcelona (Spain) E-mail: {afornells,jmmarto,elisabet,josepmg,xvilasis}@salle.url.edu WWW home page: http://www.salle.url.edu/GRSI DESMAI is a framework for helping experts in breast cancer diagnosis. It allows experts to explore digital mammographic image databases according to a certain topology criteria when they need to decide whether a sample is benign or malignant. In this way, they are provided with complementary information to enhance their interpretations and predictions. The core of the application is a SOMCBR system, which is variant of a Case-Based Reasoning system featured by organizing the case memory using a Self-Organizing Map. The article presents a strategy for improving the SOMCBR reliability thanks to the relations between cases and clusters. The approach is successfully applied in DESMAI for estimating, if it is possible, the class of the recovered mammographies. Keywords : Self-Organizing Maps; Case-Based Reasoning; Soft-Computing; Neural Networks; Breast Cancer; 1. Introduction Breast cancer is the main cause of cancer-related death in women aged 15-54. However, a high per- centage of them can be cured if they are detected in early stages. The problem is that symptoms are not evident until advanced stages, making treatments more aggressive and also less efficient. Experts are unable to do anything until it is de- tected because they do not know why the cancer ap- pears and what factors are directly related to it. For this reason, governments have made important in- vestments in massive screening programs with the aim of monitoring all the potential patients at risk and helping experts in early detection. These studies analyze the tissue composition by means of mam- mographic images, which are similar to breast x- rays. The results achieved have proved that these programs are a good methodology for prematurely detecting and removing breast cancer, increasing the survival percentage of patients. 1 Moreover, these programs have been enhanced with the development of Computer Aided Systems for helping experts in