Characterization of breast masses using texture and shape features

This work presents a study of masses classification in digitized mammograms by means of artificial neural network (ANN). The backpropagation training algorithm was used to adjust the weights of ANN. We investigated 28 features (14 shape descriptors and 14 texture features). Texture features were calculated using Haralick's analysis (spatial gray-level dependence matrices - SGLD). The images used in this study were selected from two databases: from the "Hospital de Clinicas de Ribeirao Preto", University of Sao Paulo, Brazil, and from the "Mammographic Image Analysis Society (MIAS-UK)". The process involved 118 mammograms containing lesions (68 malignant and 50 benign). The digitized regions from the mammograms were segmented using a combination of regions growing, gray-level thresholding, and morphological filtering operations. The overall result achieved an area under the ROC curve of 0.97, with an average rate of specificity and sensitivity of 93% and 94%, respectively. The best classification average rate was about 93.2%. This preliminary result indicates the usefulness of proposed schemes in characterization of mammographic masses using texture and shape features.