A fuzzy rule-based approach for characterization of mammogram masses into BI-RADS shape categories

We present new geometric shape and margin features for classifying mammogram mass lesions into BI-RADS shape categories: round, oval, lobular and irregular. According to Breast Imaging Reporting and Data System (BIRADS), masses can be differentiated using its shape, size and density, which is how radiologist visualizes the mammograms. Measuring regular and irregular shapes mathematically is found to be a difficult task, since there is no single measure available to differentiate various shapes. It is known that for mammograms, shape features are superior to Haralick and wavelet based features. Various geometrical shape and margin features have been introduced based on maximum and minimum radius of mass to classify the morphology of masses. These geometric features are found to be good in discriminating regular shapes from irregular shapes. In this paper, each mass is described by shape feature vector consists of 17 shape and margin properties. The masses are classified into 4 categories such as round, oval, lobular and irregular. Classifying masses into 4 categories is a very difficult task compared to classifying masses as benign, malignant or normal vs. abnormal. Only shape and margin characteristics can be used to discriminate these 4 categories effectively. Experiments have been conducted on mammogram images from the Digital Database for Screening Mammography (DDSM) and classified using C5.0 decision tree classifier. Total of 224 DDSM mammogram masses are considered for experiment. The C5.0 decision tree algorithm is used to generate simple rules, which can be easily implemented and used in fuzzy inference system as if…then…else statements. The rules are used to construct the generalized fuzzy membership function for classifying the masses as round, oval, lobular or irregular. Proposed approach is twice effective than existing Beamlet based features for classifying the mass as round, oval, lobular or irregular.