Differentiation of Benign and Malignant Masses in Mammogram Using 2D-Fourier-Bessel Intrinsic Band Functions and Improved Feature Space

This paper proposes a framework based on the 2D-Fourier-Bessel decomposition method (2D-FBDM) and improved feature space for the automatic diagnosis of benign and malignant masses in mammograms. For analysis purposes, a curated breast imaging subset of the digital database for screening mammography (CBIS-DDSM) is used. Haralick texture features are used to extract finesse, coarse or smoothness, and irregularities in 2D-Fourier-Bessel intrinsic band functions which are obtained by 2D-FBDM. Linear regression-based improved feature space is produced and effects on classification performance are analysed after ensambling them with old feature space. For CBIS-DDSM, the accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve obtained by the proposed framework are 99.06%, 98.48%, 99.74%, and 0.99, respectively. The mini-mammographic image analysis society database is also analyzed to show the robustness of the proposed framework.