Breast cancer is one of the major causes of death among women all over the world. An improvement of early detection and diagnosis techniques is very important for women’s quality of life. Computer-Aided Detection (CAD) systems have been used for aiding radiologists in their decision in order to solve the limitations of human observers. This paper presents a methodology for mass detection in digital mammograms. This methodology begins with segmenting Regions of Interest (ROIs) using morphological operations and automatic thresholding. Features are extracted from the ROIs and Principal Component Analysis (PCA) is applied for reducing the features dimensionality. Finally, the methodology performs classification through Neural Networks (NNs). The proposed system was tested on several mammographic images extracted from DDSM database. Results showed that the proposed methodology provided more accuracy than other compared techniques.