Abstract
Mammography is the most efficient method for early mass detection and diagnosis. This paper deals with the problem of shape features extraction in digital mammogram for mass diagnosis. We propose to combine a region and boundary features in order to ameliorate the diagnosis quality. For boundary analysis we propose to ameliorate the RDM method by using an extended approach noted XRDM We also define a new feature (IA) based on angle calculation. Based on the literature, we exploit a set of region features that are the most used and the simplest for mass description. For experiments, we use the DDSM database and some classifiers as Multilayer Perception (MLP) and K-Nearest Neighbours (KNN). Using KNN classifiers, we obtained 97.1% as sensitivity (percentage of pathological ROIs correctly classified). The results in term of specificity (percentage of non-pathological ROIs correctly classified) grew around 95.63% using MLP classifier.