Chapter 13 - Classification of Breast Density Patterns Using PNN, NFC, and SVM Classifiers

Abstract

The work introduces a computer-assisted framework to classify breast density using digitized screen film mammograms into (a) 4-class density according to BIRADS standard and (b) 2-class breast density. The “fatty” class is formed by combining mammographic images from BIRADS-I and BIRADS-II classes and similarly the “dense” class is formed by combining the mammographic images from BIRADS-III and BIRADS-IV classes. The proposed algorithm has been tested using 480 mammographic images, i.e., 120 mammographic images each belonging to BIRADS-I, BIRADS-II, BIRADS-III, and BIRADS-IV, acquired from the DDSM dataset. From each of these images, a fixed-size square ROI with dimensions 128×128 pixels is cropped from the central area of the breast where glandular ducts are prominent.