Segmentation of the breast region in mammograms using marker-controlled watershed transform

Abstract

Extraction of breast contour is a crucial first step in computer aided diagnosis of mammograms. It has the advantage of allowing the search for abnormalities to be limited to the region of the breast without undue influence from the background and also provides a reference for the alignment of views when two views are being compared. Breast region segmentation is difficult because of the tapering nature of the breast. This paper presents a marker-controlled watershed method for breast region segmentation on mammograms. It mainly consists of computation of the segmentation function and the internal and external markers. The segmentation function is determined on the smoothed morphological gradient image by morphological reconstruction, and markers are determined on basis of a developed Otsu method. Results are evaluated by comparison with manual borders drawn by radiologist for a set of 120 mammograms, comparing with other three segmentation methods. The mean ± std of the values of the area overlap metric for our method is 0.93 ± 0.03. The experimental results indicate that the breast regions extracted by our method accurately characterize the corresponding ground truth images.