Diverse and Discrimintative Features Based Breast Cancer Detection Using Digital Mammography

Mammography is one of the reliable and trustworthy methods for early diagnosis of breast carcinomas. The presence of micro calcification clusters is an important sign for the discovery of early breast carcinoma. In this work, a diverse features based breast cancer detection (DF-BrCanD) system to detect breast cancer is proposed that may be considered as a second opinion. The purpose of this work is to increase the radiologist diagnostic confidence and offer more objective evidence. We have used phylogenetic trees, statistical features and local binary patterns to generate a set of diverse and discriminative features for subsequent classification. Finally, Support Vector Machine with RBF kernel is used for the classification of mammographic images as cancerous and non-cancerous. The performance of the proposed DF-BrCanD system is analyzed using standard database for screening mammography through experimental comparison based on various performance measures. We show that the proposed DF-BrCanD system is quite effective in detecting breast carcinoma.