Characterization and Classification Algorithm for Mammography Images by means of the BIRADS Assessment Categories

According to the World Health Organization, breast cancer is the most common cancer in the world. This is a disease in which cells in the breast grow and multiply out of control. Fortunately, it can be treated and cured if it is early detected. The most widely used screening method for this disease is mammography, which has a reporting standard, called 'Breast Imaging Reporting and Data System' (BIRADS), which classifies the lesions in categories numbered from 0 to 6. The aim of this research seeks to design and implement a computer-assisted diagnosis algorithm, in order to identify and classify breast lesions using image processing techniques, as a diagnostic aid for radiologists. For this purpose, five stages were done: Image pre-processing, image segmentation (including pectoral muscle and lesions in the area) by using region-growing technique, texture and morphological features extraction and classification of the lesions. To classify the lesions, a multilayer perceptron (MLP) was used, obtaining an 74.6% of accuracy, fulfilling the objective of exceeding the accuracy of a specialized observer.