Mammogram classification using selected GLCM features and random forest classifier

Early diagnosis of breast cancer can improve the survival rate by detecting the cancer at initial stage. Mammogram is a low dose X-ray image of the breast region, used to diagnose the breast cancer at early stage. In this paper, an efficient computer added diagnosis (CAD) system is proposed, automatically detects the normal and abnormal images of mammogram. The proposed pre-processing steps include, cropping of mammograms (for avoiding the pectoral muscle, unwanted tags) and suppression of Gaussian noise. Further, gray level co-occurrence matrix (GLCM) based statistical texture feature from different distances of neighboring and angles are extracted. Furthermore, most relevant features are also examined using AdaBoost feature selection method. Finally, normal and abnormal mammograms are classified using Random forest (RF) classifier. Experiments on benchmark mammography image analysis society (MIAS) database confirm the effectiveness of this work.