Application of artificial neural networks in renal transplantation: classification of nephrotoxicity and acute cellular rejection episodes

Abstract

Complications associated with kidney transplantation and immunosuppression can be prevented or treated effectively if diagnosed in the early stages by posttransplant monitoring. One of the major problems is diseases that occur during the first year after kidney transplantation. For this purpose, we used different classifiers to predict events of nephrotoxicity versus acute cellular rejection episodes. The classifiers were evaluated according to values of sensitivity, specificity and area under ROC curves (RCA). The classifier with better accuracy rate for nephrotoxicity achieved the value of 75.68% and RCA classifier reached the accuracy of 80.89%. These results are encouraging, with rates of accuracy and error consistent with work purpose.