Breast abnormality detection incorporating breast density information based on independent components analysis

Abstract

This paper introduces an approach to breast abnormality classification which incorporates breast density information. Features are extracted by a novel technique based on Independent Component Analysis, which decomposes the selected images into sets of independent source regions and corresponding basis functions (weights). The coefficients which result from the source regions are used in turn to describe normality and abnormality. The method has been tested on the MIAS database and has high sensitivity. © 2008 Springer-Verlag Berlin Heidelberg.