Automatic Breast Boundary Segmentation of Mammograms
R. R. Nanayakkara, Y Yapa, P. B. Hevawithana, Pushpakanthie Wijekoon · 2015
Accurate breast boundary estimation and segmentation of breast tissue region from the background of the mammogram image is an important pre-processing task in computer-aided diagnosis of breast cancer. This paper presents an automated system to estimate skin-line and breast segmentatio n. The proposed method is based on automatic seed region selection, modified fast marching algorithm to propagate the seed region and automatic boundary point selection with intensit y gradient information to initial boundary estimation and morphological operators to final boundary estimation and breast t issue region segmentation. Performance of the proposed method was tested by using 136 mammogram images with all types of breast tissues taken from mini-MIAS database. The results obtained from the experimental evaluation indicate that the sensitivi ty of this algorithm is 99.2% of the ground truth breast regio n and accuracy of the segmentation is 99.0%. By analyzing the resul ts we can conclude that this system is capable of estimate th e breast boundary and segment the breast area from background for all three types of breast tissues with high accuracy lev el.