Breast Mass Segmentation in Digital Mammography Using Graph Cuts

This paper presents a novel method for the segmentation of breast masses on a mammography. Accurate segmentation is an important task for the correct detection of lesions and its characterization in computer-aided diagnosis systems. Many popular methods exist, of which most of them rely on statistical analysis. Similar to other methods, we propose a graph theoretic image segmentation technique to segment the breast masses automatically. This method consists of two main steps. First we introduce a thresholding method to obtain the rough region of the masses by eliminating all other artifacts. Then, on the basis of this rough region, the graph cuts method was applied to extract the masses from the mammography. The results were evaluated by an expert radiologist and we compared our proposed method with the level set algorithm, which shows the highest success rate. In contrast, we experiment our method on two different databases: DDSM and MiniMIAS. Experimental results show that the proposed method has the potential to detect the masses correctly and is useful for CAD systems.