Detection of Masses in Digital Mammograms using K-Means and Support Vector Machine

Abstract

Female breast cancer is a major cause of death in occidental countries. CAD/CADx systems can aid
radiologists in detection and diagnostic of lesions in mammograms. In this work, we present a methodology
to detect masses from mammograms. The K-means clustering algorithm is used to split the mammograms
in regions. Each region is then classified through a Support Vector Machine (SVM) as mass or non-mass
region. SVM is a machine-learning method, based on the principle of structural risk minimization, which
performs well when applied to data outside the training set. We use a set of textural and shape measures to
detect suspicious regions, as bening and malignant masses. Each textural measure (contrast, homogeneity,
inverse difference moment, entropy and energy) is computed through the co-ocurrence matrix technique.
The methodology obtained an accuracy of 93.11% discriminate mass from non-mass elements.