A CAD SYSTEM FOR THE DETECTION OF MAMMOGRAPHYC MICROCALCIFICATIONS BASED ON GABOR TRANSFORM

Breast cancer is the first cause of death among women in Europe. Many studies have shown that the early diagnosis is the most efficient way to fight this disease. With this in mind, Computer Aided Detection (CAD) systems are being developed to help radiologists working with mammography to assess correct diagnosis. The main goal of these systems is to direct the radiologist's attention to suspicious areas. In this paper we present the software architecture of GNN-CAD (Gabor Neural Network CAD), a CAD system for the detection and classification of breast calcifications. Our approach is based on the multiresolution space-frequency scheme to the representation of the image known as Gabor Transform. The Gabor Neural Network CAD (GNN-CAD) system works in several steps. Digitized mammograms are first preprocessed, then, the image spectral and spatial features extracted by a bank of Gabor filters at different spatial and spatial-frequency resolutions are used to train a three layers feed-forward Artificial Neural Network (ANN) to discriminate between normal pixels and pixel belonging to isolated and clustered microcalcifications. The microcalcifications thus detected are grouped in clusters and classified.