Hybrid Clustering and Texture Features in Segmentation of Breast Masses in Mammograms
Moustapha Mohamed Saleck, Abdelmajid El Moutaouakkil, Mohamed Rmili · 2018
Image segmentation plays a key role in many medical imaging applications, especially in Computer-Aided Detection (CAD) system for mammography. A good segmentation allows increasing the performance and efficiency of CAD system that enables the radiologist to conduct a clear diagnostic analysis and to make better decisions; this requires effective tools and techniques. This paper proposes a new method to extract the mass from the Region of Interest (ROI) based on texture features and Fuzzy C-Means (FCM) clustering with setting c= 2, whereas the user selects the region of interest manually. The process of clustering is applying within an appropriate range limited by the maximum of intensity and a threshold defined by the big changes in the texture features levels. The proposed method is applied to Mini-MIAS database and then its performance is compared with some explored methods. In this study, the result of overlap measure (AOM) was achieved approximately 81%.