Semi-Automatic Segmentation of Breast Masses in Mammogram Images

Breast lesion segmentation is a critical task in Computer-Aided Diagnosis (CAD) techniques for mammography, the performance of CAD system strongly depends on the results of segmentation. In this paper, we present a simple and robust approach for breast mass segmentation in digital mammograms. The proposed approach consists of three major stages. In the first stage, 2D-median filter is applied for enhancing the quality of image. In the second stage, an initial segmentation is designed based on canny and watershed algorithms; this step allows automating the process of seed point selection. In the final phase, the boundaries of tumor are extracted from Region of Interest (ROI) with high accuracy by using region growing method. The validation process of the proposed approach was achieved based on a set of 27 images from Mini-MIAS database with an average overlap of 81.3 % and the result were compared with some other mammograms segmentation methods. An experimental evaluation of this study shows that the proposed method can reliably be applied on mass segmentation in mammogram images.