Abstarct: Breast cancer is the most common cancer and the second leading cause of cancer deaths among women. Studies show that it is not nearly possible to prevent of breast cancer since its causes are not known. So, the early detection of the stage of cancer plays the key role in treatment of this cancer. The early detection and treatment of breast cancer increase the life expectancy and stimulate the patient for next visit. Now, mammography is the most common way of early detection of this cancer and the death rate has been reduced by 25%. Computer-Aided Diagnosis (CAD) systems are widely used for detection of various kinds of abnormalities in mammography images. Masses and Microcalcifications are two type important of these abnormalities in mammography images. This study aim to plan, simulate, investigate and compare CAD systems ableing automatically determine the masses im mammography images with high accuracy. In addition to the mass, the application of CAD in MCs detection are considered in short term. Our proposed method in this thesis consists of four main steps. These steps include preprocessing, extraction of the suspected areas of mass, feature extraction and classification. All steps will be discussed in detail in different chapters. In preprocessing step, some solutions for image enhancement of mammography and deletion of extra parts such as labels and pectoral muscle are present. In second step, suspicious areas are extracted from mammography image by histogram analysis. Then, Different types of intensity features, texture features and shape features are introduced. And samples of them are expressed. effective combination of texture features resulting from GLCM and GLRLM matrix will be suggested and it will result in high accuracy segregation. . in the last stage to classification of suspicious areas, three basic classifier MLP, K-NN and SVM are used. We evaluate the results of these classifiers using DDSM databaxse mammography image. Finally In order to increase the accuracy of classification, we used combination of basic classifiers output. OWA method is used for this purpose. The reported results of this method show that the accuracy of classification is increased by 4% compared with the best result in basic classifiers. Also, simultaneous use of GLCM and GLRLM features shows 2% accuracy enhancement compare with the time when each of them is used alone.