Long-Term Survival Prediction Of Liver Transplantation Using Deep Learning Techniques

Abstract

The forecasting of survival following liver transplantation is one of the greatest and most crucial areas of medical investigation. The most efficient choice of treatment for advanced liver disease is liver transplantation. Before any transplant, everyone will take their prospects of survival into account. An overview of the clinical and computational predictions for patients who have undergone liver transplants is given in this article. This research analyses multiple deep learning algorithms that can predict the survival of patients who went through liver transplants using data from the United Nations for Organ Sharing (UNOS). We considered liver transplants made in the United States of America between October 1, 1987, and June 30, 2021, using a database from the United Network for Organ Sharing (UNOS) that comprises 65535 donor-recipient pairings and 421 variables. Several methodologies, including proportional-hazards regression models and AI techniques, including Random Forest, Artificial Neural Network, Transformer, and K Nearest Neighbor were evaluated using 29 correlated features obtained through WEKA Software. All the Deep learning models were compared based on accuracy. With an accuracy of 0.89, the FT-transformer model outperformed all other models.