Morphological extreme learning machines applied to detect and classify masses in mammograms

According to theWorld Health Organization, breast cancer is the most common form of cancer in women. It is the second leading cause of death among women around the world, becoming the most fatal form of cancer. However, to detect and classify masses is a hard task even for experts. Consequently, due to medical experience, different diagnoses to an image are commonly found. Therefore, the use of a computer assisted diagnosis is important to avoid misdiagnoses. In this work, we propose Morphological Extreme Learning Machines, with hidden layer kernel based on nonlinear morphological operators of erosion and dilation. The proposed approach is evaluated using 2.796 images from IRMA database, considering fat, fibroid, dense and extremely dense tissues. Zernike Moments and Haralick texture features are used as image descriptors and the proposed model classifies the masses in benign, malignant or normal. Results shows comparison between Extreme Learning Machines using Sigmoid and Morphological Kernel, which are evaluated through classification rate and Kappa index. When using morphological kernels, the classification rate and Kappa value increases for most of cases analyzed.