Computer-aided diagnosis system for mammogram density classification

Abstract

Mammography is the widely used technique in breast cancer diagnosis. This paper presents a computer-aided diagnosis system to classify the mammograms into three different densities including fatty, glandular and dense. Mammographic density is an important factor of breast cancer risk. Higher breast densities increase the difficulty of detecting cancer in a mammogram. The accuracy of breast cancer detection depends on the breast tissue characteristics. Several texture features such as histogram, local binary pattern, gray-level co-occurrence matrix, gray-level difference matrix, gray-level run-length matrix, Gabor transform and discrete wavelet transform were extracted from the mammograms. In this work, correlation-based feature selection technique was used. The breast tissue classification based on texture features was evaluated by artificial neural network, linear discriminant, Support Vector Machine (SVM) and Naive Bayes classifier. The performance of the proposed method was examined using the Mammogram Image Analysis Society (MIAS) database. Experimental results demonstrate that the best performance was achieved by SVM yielding an accuracy of 96.11%.