Breast Mass Segmentation In Digitized Mammograms Based On Watershed And Level Set

Currently, medical image segmentation plays a critical role in medical image processing. Medical image segmentation which is the basis of further diagnosis can supply a powerful dependence for medical image processing and analysis. It's primary purpose is to separate different regions with special signification from the image and to extract the main feature data. Because of the noisy and low resolution of medical image, it becomes a classic and difficult problem in the field of medical image processing.Breast cancer is one of malignant tumor seriously damaging to women's health, and its morbidity is the first malignant tumor in the world, which has been highly aroused attention by the world. Breast mass segmentation in mammography X-ray images has extremely vital significance for the breast cancer early diagnosis and early treatment. As a result, the morbidity and mortality will be reduced and patients can gain more time for breast cancer prevention and treatment.Breast mass is a main performance of diseases on mammography X-ray images. Breast mass segmentation is often required as a significant step for monitoring and quantifying breast cancer. As medical images will be affected by many factors, it is more difficult to automatic segment breast mass on breast X-ray images. In this paper, based on the comprehensive research and in comparison to the existing breast mass segmentation technology at home and abroad, we present an efficient approach for breast mass segmentation. The proposed method is composed of two steps. Firstly, the marker controlled watershed algorithm is applied to mammograms as an initial segmentation. Then, the contour line obtained by watershed is regarded as the initial curve and a level set evolution without re-initialization is utilized for the further segmentation. This integrated approach yields a robust and precise segmentation. The effectiveness of the proposed approach is validated using extensive experiments on the MIAS and DDSM databases.