Automatic characterization of masses in mammograms

摘要: The classification of benign and malignant masses in digital mammogram is an important yet challenging step for the early detection breast cancer. This paper presents statistical measures orientation texture to classify masses. Since presence mass may change normal tissues, two types co-occurrence matrices are derived estimate joint occurrence angles oriented structures characterizing them. Haralick's 14 features then extracted from each different regions related mass. A total 444 434 scanned-film images DDSM database selected evaluate performance proposed differentiate also compared with features, obtained well-known gray-level matrix. best Az value 0.77 achieved stepwise logistic regression method feature selection, Fisher linear discriminant analysis classification, leave-one-ROI-out approach cross validation.