Mass segmentation in mammogram based on SPCNN and improved vector-CV

Abstract

Mass segmentation plays an important role in computer-aided diagnosis (CAD) system. The segmentation result seriously affects classifying mass as benign and malignant. By combining the simplified pulse coupled neural network (SPCNN) and the improved vector active contour without edge (vector-CV), a novel method of mass segmentation in mammogram is proposed in this paper. First, the parameters and termination conditions of SPCNN are obtained through mathematical analysis and the initial contour is segmented by SPCNN. Then, the vector CV model is accordingly modified to overcome the shortcomings of traditional CV model. Finally, combined with the initial contour, the improved vector-CV is used to segment the mass contour. The experiments implemented on the public digital database for screening mammography (DDSM) and the clinical images which are provided by the Center of Breast Disease of Peking University People’s Hospital indicate that the proposed method is better than the existing methods, especially when dealing with the dense breasts of Oriental female.