Digital mammogram segmentation based on normalized cuts

Breast cancer is a serious diseases, mammography can effectively help its early diagnosis and treatment. As mammograms are complex images, it is need to segment breast masses in computer-aided mammography screening system. Based on normalized cuts, this paper proposes a novel mammogram segmentation method. After preprocessing, we extracted the texture features of mammogram, and setting up the weights function with these features; then, the normalized cuts method is used to find the partitions of the mammogram. Our proposed method is verified through experiment with digital mammograms obtained from mini-MIAS database. The experimental results obtained are quite good.