Mass segmentation in mammograms for computer-aided diagnosis of breast cancer

The appearance of masses in in X-ray mammograms is one of the early signs of women breast cancer. Currently, mammography is the single most effective and reliable technique in the investigation of breast abnormalities detection such as masses. However, their detection is still a challenging problem due, to the diversity in shape, size, ambiguous margins and to the poor contrast between the cancerous areas and surrounding bright structures. This paper presents an effective approach based on mathematical morphology for detection of masses in digitized mammograms. The developed approach performs an initial step in order to remove and delete unwanted signs and radiopaque artifacts present in the background of the mammogram, and to extract the breast area. Then an enhancement process is applied to improve appearance and increase the contrast of images and to eliminate noise. Once the breast region has been found, a segmentation phase through morphological watersheds is performed for localization /detection of various types of masses in mammograms. The main advantage and motivation of this paper is the ability to detect hard masses cases in very dense mammograms. The algorithms have been evaluated on a set of 38 mammograms from MIAS dataset, shows the presence of masses, previously selected by expert radiologists. In addition, it has been compared to the manual detection marked by radiologists. The obtained results show promising performances of the proposed algorithm. Indeed, the watershed transform has demonstrated a great efficiency in the masses detection. Consequently, the developed algorithm provides “a visual aid” to radiologists in their interpreting mammograms.