Use of Artificial Neural Networks in Improving Renal Transplantation Outcomes

Abstract

Recent advances in renal transplantation, including the matching of major histocompatibility complex or new immunosuppressants, have improved 1-year survival of cadaver kidney grafts to more than 85%. Further optimization of kidney transplant outcomes is necessary to enhance both the graft survival time and the quality of life. Techniques derived from the artificial intelligence enable better prediction of graft outcomes by using donor and recipient data. The authors used an artificial neural network (ANN) to model kidney graft rejection and trained it with data on 1542 kidney transplants. The ANN correctly predicted 85% of successful and 72% of failed transplants. Also, ANN correctly predicted the type of rejection (hyperacute, acute, subacute, and chronic) for approximately 60% of the failed transplants. These results indicate that the ANN-based approach is useful for prediction of both the general outcomes of kidney transplants and the prediction of the type of rejection.