Breast Epithelial Duct Region Segmentation Using Intuitionistic Fuzzy Based Multi-Texture Image Map

Currently histology breast cancer tissue interpretation through microscope is a manual procedure. Inter-intra-observer variability in breast cancer diagnosis by expert pathologist can be improved by minimizing it through automatic objective analysis. To achieve this purpose, an Atanassov's Intuitionistic Fuzzy Set (AIFS) based multi-texture image clustering technique is proposed in this paper to segment duct epithelial arrangement present in the breast biopsy images. Main contribution of this paper includes development of an AIFS color texture map to capture breast duct region from digitized histology images. Further, this multi-feature texture image representation is used to formulate new AIFS affinity matrix. Normalized cuts (NCUT) method using this new affinity matrix automatically segments duct epithelium nuclear arrangement. The algorithm is applied on digitized Non-Malignant (NM) and Malignant (M) breast biopsy tissue images to automatically detect epithelial nuclear region and is evaluated quantitatively as well as qualitatively. The algorithm segments epithelial duct from the breast histology images at average segmentation accuracy of 88.54% and 77.05% respectively. Qualitative and quantitative evaluation of segmentation results for proposed algorithm demonstrates the efficiency of algorithm in accurate segmentation of breast duct regions.