Prediction of Kidney Graft Rejection Using Artificial Neural Network

Area of Science:

• Nephrology
• Transplant Surgery
• Artificial Intelligence in Medicine

Background:

• Kidney transplantation is the optimal treatment for end-stage renal disease.
• Identifying predictors of kidney graft rejection remains challenging due to inconsistent findings in previous studies.
• Accurate prediction of graft rejection is crucial for improving patient outcomes.

Purpose of the Study:

• To identify prognostic factors for kidney transplant rejection using artificial neural network (ANN).
• To compare the predictive performance of ANN with logistic regression (LR) for graft rejection.

Main Methods:

• Retrospective analysis of 378 kidney transplant recipients in Hamadan, Iran (1994-2011).
• Application of ANN to identify significant risk factors for chronic, irreversible graft rejection.
• Comparison of ANN model performance against LR using accuracy and ROC curve analysis.

Main Results:

• ANN identified recipient age, creatinine, cold ischemic time, and hemoglobin level as key prognostic factors.
• The ANN model demonstrated superior predictive accuracy (0.75) compared to LR (0.55).
• ANN achieved a higher area under the ROC curve (0.88) than LR (0.75), indicating better discrimination.

Conclusions:

• Artificial neural networks significantly outperform logistic regression in predicting kidney transplant failure.
• ANN is a promising tool for predicting graft failure, potentially enhancing patient survival and quality of life.
• Further research into ANN for predicting clinical outcomes in transplantation is warranted.