A Neutrosophic Approach for Breast Mass Detection in digital Mammogram Images

Abstract

Breast cancer is the most common type of cancer that affect women and yet there is no substantive cure. On early diagnosis of breast lesions with this size using mammogram would help the affected tumor patients have a good survival rate. Detecting cancer cells is a challenging task and hence we propose a single valued neutrosophic approach for segmenting breast tumor lesions in mammogram images. In this neutrosophic approach, the intensity values are represented as truth, indeterminacy and falsity membership in where the indeterminacy are noise, and the breast regions are the true and falsity values. The computations of these memberships are then pre-processed using α-mean and β-enhancement to minimize the indeterminacy. Finally, gamma clustering techniques are used to localize and segment the tumor from the background breast tissue region. Experiments are conducted on publicly available datasets and the proposed method is evaluated and achieves a better segmentation performance.