Prediction of Graft Survival Using Artificial Neural Network (ANN), and Bayesian Belief Network (BBN): A Comparative Study Graft Survival Prediction Using Machine Learning: A Comparative Study

Abstract

Objective: The aim of this study was to predict graft survival using machine learning prediction techniques and the involved decision making.

Design: Prediction of graft survival post-transplant using machine learning algorithms like Artificial Neural Network (ANN) (Single and Multi-layer networks), and Bayesian Belief Network (BBN).

Setting: Recipient and donor with characteristics of age, sex and Glomerular Filtration Rate (GFR) and the follow-up of probability of survival one year after transplantation (n=40).

Main outcome measures: The Data include simulation from donor, recipient characteristics of single centre with factors age, sex, GFR and probability of survival collected particularly with the follow-up after the first year of transplant.

Results: The ANN and BBN were modelled in Python. The probability of survival post-transplant is predicted, and accuracy measured using Root Mean Square Error (RMSE). The results for the methods were compared and efficacy and ease of use are discussed.

Conclusion: The decision making in the organ transplantation involving the patients and doctors consists of mainly involve improving the graft survival and hence prediction becomes important. The developed models can be used to predict the transplant and aid as decision support system for decision regarding matching and allocation.