Detection of Distrustful Masses in Digital Mammograms

An algorithm to detect the distrustful masses from mammographic images is presented. Extensive experiments have been performed with images of craniocaudal and mediolateral views obtained from the MIAS database. Results are compared with existing CAD methods that uses classification methods such as artificial neural networks, fuzzy neural networks and decision trees. The experimental study shows that the use of AdaBoost with SVM as weak classifier gives better results among other classification methods. Based on the subjective evaluation, this study can improve visual quality of images and segmentation of masses might aid clinicians in diagnosis.