Prognosis prediction of human breast cancer by integrating deep neural network and support vector machine: Supervised feature extraction and classification for breast cancer prognosis prediction

Breast cancer is the second most common cancer, which accounts for about 16% of all female cancers and 18.2% of all cancer deaths worldwide including both males and females. Therefore, accurate prognosis prediction of breast cancer is one of the most crucial works in follow-up treatment. In this work, we introduced a novel deep learning (DL) method named D-SVM for the prediction of human breast cancer prognosis. D-SVM was a supervised feature extraction and classification method integrating deep neural network and traditional support vector machine (SVM) classifier. This novel algorithm effectively learned hierarchical and abstract representation from raw input data and successfully integrated traditional classification method. Our experimental results clearly indicated that the proposed method could enable us to train a model that yielded reasonably good performance than traditional classification approaches on a small-scale gene expression dataset for breast cancer.

I. Introduction

Breast cancer is the second most common disease in females, and a primary leading cause of cancer deaths worldwide [1]. The severity of breast cancer can be divided into 5 stages (0-IV), which describes the extent of the breast cancer within patients' body [2]. The high mortality rate of breast cancer is mainly caused by initial diagnosis at the advanced stage. According to the statistics approved by Cancer. Net, most of breast cancer patients are diagnosed at the advanced stage, for which the 5-year survival rate is only about 26%. On the contrary, the 5-year survival rate of patients diagnosed at the early stage is up to 90%. Thus, the molecular mechanisms that intermediate longer survival time in breast cancer patients at the advanced stage are a hot spot in current research. However, the complexity of breast cancer and its significantly varied clinical outcomes now make it difficult to predict and treat [3]. Thus, being able to predict cancer prognosis more accurately not only assist breast cancer patients know about their life expectancy, but also assist physicians make informed decisions and further guide clinical therapy.