A Random-Walk Based Breast Tumors Segmentation Algorithm for Mammograms

Mammography is a commonly performed imaging study for screening breast cancer. One of the crucial elements of the mammogram processing is an accurate segmentation of the breast tumor region from a mammogram because it directly influences the subsequent analyzing accuracy and the processing speed of the mammogram. The goal of this paper is to propose an accurate and efficient algorithm of breast tumor extraction from the medio-lateral oblique (MLO) mammograms. The proposed algorithm adapts the modified gradient vector flow (MGVF) snake to determine the breast region from a mammogram image, and uses Otsu thresholding and multiple regression analysis to delete the pectoral muscle from the breast region. It further utilizes upper outlier detection and texture complexity analysis to segment the initial breast tumor regions, and finally, segments the final breast tumor image from the initial breast tumor regions by using random walk scheme. The proposed algorithm is tested on the digital mammograms from the Mammogram Image Analysis Society (MIAS) database. The experimental results show that the tumor extracted by the presented algorithm approximate accurately to the actual tumor regions confirmed on the biopsy results of MIAS.