Deep texture representation for breast mass classification
Sawsen Boudraa, Ahlem Melouah, Hayet Farida Merouani · 2018
The texture is the most used feature for breast mass classification. In order to look deeper into this feature, this work proposes an approach based on Local Binary Pattern and second-order features calculation. Firstly, we applied Local Binary Pattern transformation on the masse region. Secondly, we extract Gray Level Co-occurrence Matrix (GLCM) features from the transformation result. We tested the proposed approach on MIAS database by using six different classifiers. All classifiers give the highest accuracy compared to the classic classification method.