Thesis (M.Sc.) - Cairo University - Faculty of Graduate Studies For Statistical Research - Department of Information Systems and Technology

Breast cancer is one of the major causes of loss of life among women in the entire world. The detection of breast cancer in initial stages can keep many women's lives. However, almost of the early detecting systems are expensive and highly complex and takes a long time in processing; that make it unsuitable for developing countries. On the other hand, the digital mammography is considered one of the most significant diagnostic methods for breast cancer tumors. However, detecting the abnormality from a mammogram is a challenging task for radiologist. Many Computer Aided detection and Diagnosis (CAD) systems are designed to assist radiologists providing an automatic detection and diagnosing for cancer in mammograms. Region of interest (ROI) segmentation is a challenging task and an important critical process in the development of computer aided detection (CAD) system for breast cancer. This thesis proposes a computer aided detection and diagnosis (CADx) approach for breast cancer detection and diagnosis from digital mammography based on three different machine learning algorithms. The performance of the three systems is evaluated to recommend the best compromise between cost and accuracy. The proposed CADx system composed of segmentation step followed by Support Vector Machine (SVM) for automatically classifying the mammograms. We perform pre-processing enhancement for the image (i.e., label removal, extract the breast region from the background and remove pectoral muscle existing in the mammogram images) towards increasing the accuracy of our CADx system

Issued also as CD