Classification of Mammogram Images Using Hybrid Features

Abstract

Breast cancer is the most common cancer in cities of India, and second most newlinecommon in the rural areas. Creating awareness on breast cancer and early detection is newlinethe only way to reduce the mortality. Despite the fact, not all general hospitals have the newlinefacilities to diagnose breast cancer through mammograms. Awareness and newlineunderstanding of this disease make more women come forward for early detection of newlinebreast cancer by taking mammogram. Delay in diagnosing the breast cancer for a long newlinetime may increase the possibility of the cancer spreading. Therefore a computerized newlinebreast cancer diagnosis has been developed to start the treatment early and reduce the newlinedeath rate. Physicians need a reliable diagnostic procedure to detect the cancerous newlinetumors at its early stage because early detection is the key for breast cancer prognosis. newlineAlso the tumors may be normal or cancerous. But generally it is very difficult to detect newlineand distinguish tumors, even by the experts. Hence automation of diagnostic system is newlineneeded for diagnosing tumors. Digital mammography plays an important role in newlineComputer Aided Detection of Breast Cancer. This work is an effort to provide an newlineautomated method for detection of breast cancer with classification model and newlineexperimental results so as to detect the tumors at its early stage and make better newlinedecisions. The mammogram images used in this study are taken from the Mini- newlineMammographic Image Analysis Society (Mini-MIAS), Digital Database for Screening newlineMammography (DDSM) digital databases and local city based health centers. newlineThere are five distinct parts in this research work. First part, the review of newlineliterature deals with the survey of existing methods that have been developed for newlineclassification of mammogram images. Second part conducts a study to analyze the newlineperformance of various feature selection algorithms and third part proposes a new newlinealgorithm based on a hybrid approach which identifies the most relevant features used newlinefor classifying the mammogram images. Fourth part dea