A Hybrid Feature Extraction Approach for Mammogram Mass Analysis

This paper focuses on characterizing the spatial structure of mammogram masses using various hybrid features. According to Breast Imaging Reporting and Data System (BIRADS) spatial benign and malignant masses can be differentiated using its shape, size and density, which is how radiologist visualizes the mammograms. Based on BIRADS mass shape characteristics, benign masses tend to have round, oval, lobular in shape and malignant masses are lobular or irregular in shape. Measuring regular and irregular shapes mathematically is found to be a difficult task, since there is no single measure to differentiate various shapes. Various new geometrical shape and margin features like eccentricity, Elongatedness, dispersion, circularity were introduced to characterize the morphology of masses. This approach uses 20 hybrid features like 17 shape/margin/texture features and DDSM database descriptors like density, assessment, subtly. A Univariate Analysis Of Variance (ANOVA) Discriminant Analysis (DA) classifier is used for classifying spatial structure of mass either as benign or malignant. The study shows the contribution of shape and DDSM descriptors in discriminating benign and malignant masses. Experiments have been conducted on benchmark Digital Database for Screening Mammography (DDSM) database. The experimental results shows that proposed approach performs superior using hybrid features compared to unimodal features for mammogram mass classification.