According to the World Health Organization, breast cancer is the main causeof cancer death among adult women in the world. Although breast cancer occursindiscriminately in countries with several degrees of social and economicdevelopment, among developing and underdevelopment countries mortality ratesare still high, due to low availability of early detection technologies. Fromthe clinical point of view, mammography is still the most effective diagnostictechnology, given the wide diffusion of the use and interpretation of theseimages. Herein this work we propose a method to detect and classifymammographic lesions using the regions of interest of images. Our proposalconsists in decomposing each image using multi-resolution wavelets. Zernikemoments are extracted from each wavelet component. Using this approach we cancombine both texture and shape features, which can be applied both to thedetection and classification of mammary lesions. We used 355 images of fattybreast tissue of IRMA database, with 233 normal instances (no lesion), 72benign, and 83 malignant cases. Classification was performed by using SVM andELM networks with modified kernels, in order to optimize accuracy rates,reaching 94.11%. Considering both accuracy rates and training times, we definedthe ration between average percentage accuracy and average training time in areverse order. Our proposal was 50 times higher than the ratio obtained usingthe best method of the state-of-the-art. As our proposed model can combine highaccuracy rate with low learning time, whenever a new data is received, our workwill be able to save a lot of time, hours, in learning process in relation tothe best method of the state-of-the-art.
展开▼
