Application Of Cox Proportional Hazard Model And Back-propagation Neural Network For Prognostic Analysis Of Liver Transplantation Recipients

Objectives: To probe into prognostic influence factors of liver transplantation recipients. To compare predictive capability among model for end-stage liver disease, Cox proportional hazard model and back-propagation neural network. Back-propagation neural network was applied to forecast survival time of liver transplantation recipients and provide a new idea for forecasting survival time.Methods: Cox proportional hazard model was applied to probe into prognostic influence factors of liver transplantation recipients. Using area under ROC curve and concordance index as the evaluating indexes, we could judge predictive capability among the three models. To compare predictive survival time with actual survival time, we could estimate the capability of back-propagation neural network on forecasting survival time.Results: Main prognostic influence factors related to benign liver transplantation recipients are UREA and APTT. Main prognostic influence factors related to malign liver transplantation recipients are ALK, AFP, Na and NODE1. Based on benign liver transplantation data and malign liver transplantation data, we drew a conclusion that the predictive capability of back-propagation neural network was better than that of Cox proportional hazard model, and the predictive capability of Cox proportional hazard model was better than that of model for end-stage liver disease, and there was no...