XBPF: An extensible breast cancer prognosis framework for predicting susceptibility, recurrence and survivability

Abstract

Breast cancer is the second most lethal type of cancer causing death of woman. As a thumb rule prevention is better than cure. Prevention is possible with life style changes and healthy habits. It is also important to have early detection of it to prevent death. Many researchers contributed towards early detection, prognosis and better treatment of breast cancer in the last two decades causing decline of mortality rate. However, the breast cancer problem is still alarming and needs further research in the area of betterment of detection and prediction besides methods for treating it. Breast cancer prognosis is the holistic approach that covers three important aspects of research including prediction of susceptibility, recurrence and survivability. In this paper we propose an Extensible Breast Cancer Prognosis Framework (XBPF) for breast cancer prognosis which includes susceptibility or risk assessment, recurrence or redevelopment of the cancer after resolution, and survivability. We proposed a representative feature subset selection (RFSS) algorithm that is used along with SVM to improve efficiency in prognosis. SEER dataset is used to have experiments. A prototype is built to demonstrate proof of the concept. Our empirical study revealed that the framework is useful in prognosis of breast cancer instead of focusing on a particular aspect like susceptibility, survivability and recurrence individually. SVM-RFSS has shown significant performance improvement over state of the art prognosis methods.