Classification of Mammograms by the Breast Composition

Breast cancer produces a high rate of mortality worldwide. Early diagnosis is essential for treatment, however it is difficult to analyse high density breast tissues. Computer-aided diagnosis systems have been proposed to classify the density of mammograms, having as a major challenge to define the features that better represent the images to be classified. In this study, besides comparing them to other techniques, different texture descriptors for the representation of breast tissue density on mammograms are analyzed. In the experiments, 320 mammograms from MIAS Database are used, and the highest accuracy obtained is 77.18% in a 10-fold cross-validation scheme.