Article Preview
Top1. Introduction
Breast cancer is one of the most frequent and lethal diseases that affects women worldwide (Ferlay et al., 2015). More than 2 million new cases are recorded each year, according to the World Health Organization (WHO). At present, breast cancer treatment and evaluation of risk factors depend on the disease’s diagnosis. Bloom-Richardson grading system (Elston & Ellis, 1991) describes the detection of malignant breast cancer based on the scoring of three morphological features i) the number of mitotic elements in most active areas ii) tumor tubule formation iii) nuclear pleomorphism. This grading scheme, on the other hand, is carried out by a pathologist who examines a biopsy tissue sample under a microscope. (Robbins et al., 1995). As a result, the researcher uses image analysis approaches to try to improve this detection procedure and reduce the effort. (Meijer et al., 1997). Also, the use of Computer-Aided Diagnosis (CAD) systems brings a new era of disease diagnosis since they improve the decision-making based on the analysis of the concerned diseases (Gurcan et al., 2009; Irshad et al., 2014). The analysis of digital pathology images through image processing or CAD-based system get the attention of clinical practitioner and researchers by reducing time and precision due to the faster and reproducible image analysis. The complex nature of histopathological images presents important challenges for manual inspection, which may lead to large inter-observer variations. For that, CAD systems can substantially reduce the false-positive results delivering a more accurate diagnosis of diseases (Irshad et al., 2014). In CAD-based diagnosis, image segmentation is one of the most fundamental steps. A good segmentation leads disease detection easier and more perfect. The proper segmentation method considers several quantitative variables such as texture, size, color, and other imagenomics (Gurcan et al., 2009; Irshad et al., 2014). However, these segmentation methods confront many issues such as noise, contrast among background and foreground, cell variations, shape, size, and intracellular intensity heterogeneity (Gurcan et al., 2009; Irshad et al., 2014).