Kernel based classifiers fusion with features diversity for breast masses classification

This paper investigated a computer-aided diagnosis for breast mass classification by mammography examination using complementarity existing between features and classifiers. It is concerned with the design and development of an automatic mass classification of mammograms. The proposed method consists of three stages: segmentation, feature extraction and classification. In classification phase, kernel based classifiers combination is a current active paradigm in the field of machine learning. It takes benefit of classification fusion algorithms. The combination of Kernel-based classifiers was proposed as a research way allowing recognition reliability by using diversity which can be exist between classifiers. The proposed scheme is based on combination of support vector machine classifiers. Each one is associated with an homogenous family of features (Hu moments; central moments, Haralick moment. Diversity criteria between features are adopted in this study to ensure best performance. Our experiments demonstrated that developed system using (DDSM) database achieve very encouraging results.