Mammography is the most common modality for screening breast cancer. In this paper, a computer aided system is introduced to diagnose benignity and malignancy of masses. In the first step of the proposed method, masses are prepared for segmentation using a noise reduction and contrast enhancement technique. Afterward, a region of interest is segmented using a new adaptive region growing algorithm, and boundary and texture features are extracted to form its feature vector. Consequently, a new robust architecture is proposed to combine weak and strong classifiers to classify masses. Finally, the proposed mass diagnosis system was also tested on mini-MIAS and DDSM databases. The accuracy of the obtained results is 93% in the database of MIAS and 90% in the database of DDSM. The obtained results indicate that the proposed system can compete with the stateof- the-art methods in terms of accuracy.