Extended Mammogram Classification From Textural Features

"The efficient analysis of digital mammograms has an important role in the early detection breast cancer and can lead to a higher percentage recovery. This paper presents extended computer-aided diagnosis system for classification into three classes (normal, benign malignant). The performance is evaluated two different mammogram databases (MIAS DDSM) order assess its robustness. We discuss changes required system, particularly at level image preprocessing feature extraction. Computational experiments are performed based on methods extraction, selection classification. results indicate accuracy 66.95% MIAS dataset 54.1% DDSM obtained using genetic algorithm Random Forest Keywords: Breast detection, Mammogram classification, GLRLM, Feature selection, Forests, MIAS, DDSM. "