Mass Segmentation And Feature Extraction Of Mammographic Breast Cancer Images In Computer Aided Diagnosis System CADs

Abstract

Breast cancer cases are rising since last few years in India as well as world and it is reported as second leading cause of death in developed countries and third leading cause of death in developing countries. Breast cancer is a kind of cancer present in breast of women. It can be categorized into benign and malignant cells or masses. Benign masses are generally found in shapes like circular or oval masses but malignant masses are very irregular or more speculated in shape. On the basis of these structural properties of the masses cancer is identified. So the masses are abnormality present in the breast and for detecting these masses mammography is used. To support mammography procedure Computer-aided-diagnosis system (CADs) is used. The CAD system is automated and semi automated tools, developed to assist radiologist in the detection of masses and evaluation of mammographic images. It encloses development of database for mammographic images, diagnosis of breast cancer with help of radiologist and determination of cancer stage by their mass size. newlineThe CAD system combines pre processing, image segmentation and post processing as main stages for the analysis of medical images and its diagnosis. Mammographic image is affected by some kind of noise that must be removed before performing image analysis. Segmented mammographic image contains the suspected region which is subjected to feature extraction process. Masses are extracted from mammographic image by the segmentation methods and their various statistical features size, area, shape and other parameters are calculated and also classification have done whether the masses are benign or malignant. newlineNew approaches or different combination of techniques can be used in order to create better algorithms for more efficient diagnosis of breast cancer. Different enhancement methods were tested on the GRSDB database and found that the Gray Level Clustering-Contrast Enhancement (GLC-CE) (and#945;=0.6) methods have better PSNR and CNR values. Then different segmentation methods were