“…The detailed description of each feature definition and computational method has been reported in our previous study (Tan et al , 2014b). Basically, these features are divided into the following categories: (1) the statistical pixel value (or intensity) based features (i.e., mean, standard deviation, skewness, and kurtosis of the pixel gray values) (Wang et al , 2011), (2) fractal dimension based features to quantitatively assess breast tissue composition and mammographic density (Chang et al , 2002), (3) gray level run length based texture features computed from the gray level resolution reduced images (from 4095 to 256 gray levels) along four different directions (Tang, 1998), (4) the first-order statistics of the x -axis and y -axis histogram (cumulative projection) based features presented by Tzikopoulos et al (Tzikopoulos et al , 2011), (5) the gray level co-occurrence matrix (GLCM) based texture features proposed by Haralick et al (Haralick et al , 1973), (6) segmented breast area size, and (7) a percentage density (PD) measure (Byng et al , 1994).…”