Modeling and Predicting the Survival of Breast Cancer Patients via Deep Neural Networks and Bayesian Algorithm

Abstract

Introduction: Due to breast cancer's prevalence as the second leading cause of cancer death worldwide, accurate survival prediction models are crucial. This study aimed to use an optimized deep neural network to predict breast cancer patient survival.Material and Methods: The present study is an analytical study. The information utilized in this research is derived from the METABRIC database, associated with the molecular categorization of breast cancer patients, from the International Federation of Breast Cancer's Molecular Taxonomy Data. The total number of patients studied is 1981. Of these, 888 patients were under care until their death, and the remaining patients withdrew from the study during its course. In this database, 22 clinical features of patients have been considered, which includes a total of 6 quantitative features and 16 qualitative features. A deep neural network model called the optimized DeepHit is used to predict survival. The optimal parameters for specific variables of the neural network are obtained by the Bayesian algorithm.Results: The optimized model has achieved the criterion of c_index = 0.748, which is a criterion for measuring the capability of survival analysis models.Conclusion: The proposed model was compared with previous models using real and synthetic datasets. The results show that the optimized DeepHit achieved significantly better performance and statistically significant improvements over previous methods.

Keywords

• breast neoplasms
• deep learning
• health information systems
• medical informatics
• prognosis
• survival analysis