Extract roi from mammogram images using morphological and k-means clustering

Abstract

Cancer is leading cause of death in worldwide. Cancer cells damages all cells surrounded by it and hence covers the complete area of the body. Amongst women breast cancer is most common deadly disease as compared to men. In any type of cancer early-stage detection plays vital role as it can save the patient’s life. If cancer is diagnosed early by using breast self-examination (BSE) and clinical breast examination (CBE) at 40-49 years of age, the survival rate of breast cancer reaches 100 per cent. New strategies named CAD (computer-assisted diagnosis) programs for early detection utilizing multiple mammogram datasets, such as mini-mias, DDSM, etc., CAD (Computer Aided Diagnosis) systems are mostly used for the second opinion for radiologists. Many researchers are already developing different CAD systems. Different databases are available for the researchers to study and detect the cancerous tumors. In this paper Mini-Mias database is used. Mammogram images are low contrast and may contain noise. Filtering or pre-processing is required to remove any noises present in the image, here Weiner filter is used. Two methodologies are compared here. First using morphological operated segmentation and second is by using k-means clustering algorithm.