A Study On Breast Mass Segmentation In Mammograms Based On Active Contour Model

Breast cancer remains a leading cause of cancer deaths among women in many parts of the world. Mammography is an effective technology for early detection and diagnosis of breast cancer. Research in past years has shown that computer-aided diagnosis (CAD) technology can improve the efficiency and accuracy in mammographic breast cancer detection/diagnosis. As an important step in mass detection and/or classification, mass segmentation plays a significant role in many CAD systems.This thesis firstly describes the status of breast cancer and the breast X-ray image CAD systems. A brief review of mass segmentation is then provided. Considering the characteristics of mammographic images, we proposed to explore the application of active contour model for breast mass segmentation in this thesis study.There are two main types of active contour models, i.e. paralmetric active contour model (P-AC) and geometric active model (G-AC). Parametric active contour model detects image edge by minimizing the contour energy function. Traditional parametric active contour models, such as Gaussian force and gradient vector flow ACs, can not deal with effectively the segmentation of the medical image with a week border. In this study we propose a novel scheme for segmentation of breast mass, which is based on gradient vector flow (GVF) snake and multi-scale analysis. By combining the gaussian pyramid method and GVF Snake model, the proposed method can effectively improve the segmentation performance with less influences of initial contour selection and and the image noise level.Because the parametric active contour model has several the shortcomings in segmentation, such as the weekness on topology change, the second part of thesis research is directed to the study of geometric active contour model for extraction of breast mass contour. Based on the theory of curve evolution and level set method, a simplified Mumford-Shah model (CV model) is introduced. Because of the gray-scale heterogeneity of mammographic image, segmentation results of CV model are often not satisfactory. We proposed a new approach to breast mass segmentation in this study, which is based on level set method and multi-scale analysis. It can make full use of both global and local information in mass segmentation. It also has good stability and topology change capacity. The experimental results demonstrated that the proposed method has higher accuracy and robustness than the conventional ones.