An Efficient Abnormality Classification for Mammogram Images

In this work we focus on developing an accurate and lightweight automated diagnostic system for classifying abnormalities in breast cancer mammogram images using the Local Binary Patterns (LBP) feature extraction method at the heart of a machine learning model. We empirically show the success of the approach in recognising the presence or absence of abnormality(ies) with high predictive performance in terms of accuracy, precision, recall, and F1-score, and against other features extraction methods employed within different a number of classifiers. The reported results show a minimum of 13% performance gap in favour of our selected approach to the closest model. Our approach exceeds 90% in most of the models when applied over DDSM, the breast cancer mammography benchmark dataset.