Breast cancer is the most frequently diagnosed cancer in women all over the world. Patents have shown that cancer is the fifth cause of death worldwide, among other leading cancers, such as lung cancer, stomach cancer, liver cancer and colon cancer. In this paper, we present a method for extraction of the morphological and texture information and attribute selection for mass classification using the fusion of information from CC and MLO views. In the extraction stage, the wavelets coefficients and the singular value decomposition (SVD) technique were applied to reduce the number of texture attributes. From the segmented mass regions, we construct the mass morphological features set. The application of analysis of variance (ANOVA) also contributes to the reduction of the textural and morphological information. In the final stage, we used the Random Forest and Support Vector Machine algorithms for classifying masses in mammograms. The overall performances of the methods were evaluated by means of the area under the ROC curve (AUC). The experiments showed that the fusion of information of views contributed to increase of values of AUC. These results demonstrate that the proposed fusion of information and combination of descriptors contribute in the classification of breast lesions.
Keywords: ANOVA, mammograms, morphological, random forest, SVM, texture, wavelet.