AUTOMATIC SYSTEM FOR THE ANALYSIS AND THE DISCRIMINATION OF BREAST NODULES IN ULTRASOUND IMAGING

The actual diagnostic protocol for breast lesions includes the use of ultrasound imaging in order to determine whether the lesion is malignant or benign. The low specificity associated to the ecography induces the development of automatic systems for the parameters extraction and classification of nodules. This work proposes an algorithm able to determine some of the BIRAD features such as lesion shape, axes ratio (longitudinal and latitudinal) and the presence of a hyperechogenic halo, which is a symptom of benignity. These features were then used for the discrimination of 85 available nodules revealed by ultrasound imaging through an Artificial Neural Network (ANN). Finally, a comparison among the classifier's performance with 3 different combinations of the same parameters was realized.