Mass Segmentation in Digital Mammograms

Digital mammograms are among the most difficult medical images to read, because of the differences in the types of tissues and their low contrasts. This paper proposes a computer aided diagnostic system for mammographic mass detection that can distinguish between tumorous and healthy tissue among various parenchymal tissue patterns. This method consists in extraction of regions of interest, noise elimination, global contrast improvement, combined segmentation, and rule-based classification. The evaluation of the proposed methodology is carried out on Mammography Image Analysis Society (MIAS) dataset. The achieved results increased the detection accuracy of the lesions and reduced the number of false diagnoses of mammograms.