A Texton-Based Approach for the Classification of Benign and Malignant Masses in Mammograms

Abstract

Classification of benign and malignant masses in mammograms is a complex task due to the appearance similarities in both classes. Thus, classification of masses in mammograms is considered an important step in the development of current Computer Aided Diagnosis (CAD) systems. In this paper, we present a way to classify masses without the need for segmentation. A supervised texton-based approach is developed using filter bank responses. Subsequently, a Support Vector Machine (SVM) classifier is used to classify the images. We evaluated the results on a subset of publicly available dataset (DDSM) and obtained classification accuracy of 96% which is comparable to the state-of-the-art techniques developed for the task of mammographic mass classification.