Extracting region of interest in full-field digital mammogram with glcm and grow cut region based segmentation method.

Abstract

Early breast cancer detection can be done by analyzing mammograms and due to that breast cancer deaths can be decreased. Many methodologies can be followed to detect breast cancers like mammography, MRI, Ultrasound etc. Mammography is a low-cost and simple way of finding breast cancers. Because of early detection, the survival rate of the patient increases and it can be performed easily using digital mammograms because they are easy to capture and manipulate the images, so the abnormalities can be seen more easily. Sometimes, manual reading with expert radiologist will result in misdiagnosis. To avoid this, many computer-aided detection methods have been developed to identify the masses and micro-calcification. Before identifying, the segmentation of mammogram is a vital part and in this paper, the segmentation of mammogram can be done using region growing technique, by calculating five features (mean, dissimilarity, sum average, sum variance and correlation) using Gray Level Co-occurrence Matrix (GLCM). Then by fixing up a threshold value and grow cut method is followed to track the closed region after automated seed point detection. The features extracted are compared with ground truth values to confirm the effectiveness of the proposed method. This results of proposed method show improved performance than existing methods.