Applying Artificial Intelligence to Survival Prediction of Hepatocellular Carcinoma Patients

Abstract

Cancer, a disease that has gradually attracted attention in the world now, it has even become one of the main causes of death. Among them, liver cancer occurs in the liver or deadly tumors starting from the liver. According to the World Cancer Report (2014), the liver cancer is primary the cancer that the occurrence of 6% ranked second highest, and the death rate is 9% ranked sixth. So the liver cancer has become the target of academic research and discussion. If we can find out the key factors that affect the death of liver cancer when identifying liver cancer, it can improve the survival prediction of patients with liver cancer, and it will bring more effective treatment and confidence to the disease. In this paper, we chose a data that a real clinically diagnosed HCC patient was collected from a University of Coimbra and Coimbra Hospital in Portugal, and we separated the data into testing and training to predict the death of HCC and find out the key factors from the prediction model. The prediction model includes Decision Tree (DT), Support vector machine (SVM), and Logistic Regression (LR). The results of this paper showed that the G-means of the three modeling methods are 0.76 (LR), 0.72 (DT), and 0.68 (SVM). The best performance is logistic regression (LR), and find out the key factors that affect the survival rate of HCC include Aspartate transaminase (U / L), Age at diagnosis, and Alkaline phosphatase (U / L).