Detection and Classification of Breast Cancers from Mammography Images using Medical Image Processing and Pattern Recognition Methods

— In this paper, we presented extracted various features based on the geometry of the breast cancer masses for classification. First, after receiving the digital breast mammogram images from the DDSM database, we first consider the preprocessing of the images (including noise reduction, filter of image, image cropping and etc.). Then, using image processing techniques, an algorithm is developed that can extract the cancer mass completely intelligent from other healthy parts of the breast texture and display it completely apart. In this method, by applying the threshold technique to each of the images, we extract the boundaries of the cancer masses and then extract the properties of the extracted masses. Then to generate a feature vector, we used 18 features that they were extracted from each identified image mass. In this paper, in addition to determine the benign or malignant masses of the extracted masses, the masses were classified according to their various forms, such as circular, elliptical and irregular. The classification of the MLP neural network, the fuzzy classifier of TSK, and statistical based algorithm Bayesian classification for determining the final classification of the masses extracted from breast images were used in this paper. The results of this paper show the simplicity and high accuracy of the proposed algorithm.