Pectoral Muscle Segmentation on Digital Mammograms by Nonlinear Diffusion Filtering

Abstract

The pectoral muscle represents a predominant density region in the most medio-lateral oblique (MLO) views of mammograms. Presence of pectoral muscle in the mammogram may affect the resulting of image processing and could bias the detection procedures. So during analysis, the pectoral muscle should preferably be excluded from processing. We proposed a new method for the identification of the pectoral muscle in MLO mammograms based on nonlinear diffusion algorithm which is an edge preserving smoother. The proposed method is applied to 90 mammograms from Mammography Image Analysis Society (MIAS) database. We compared our results by those recognized by two expert radiologists. To evaluate the accuracy of proposed method, HDM (Hausdorff distance measure) and MAEDM (mean of absolute error distance measure) were used. Then we compared our results by two other pectoral muscle segmentation methods proposed by Karssemeijer and Ferrari. The first is based on Hough-transform and the second is based on Gabor-filters. Our proposed algorithm shows superior results in comparison.