Published in:
01-10-2016 | Original Article
Predicting the survival of graft following liver transplantation using a nonlinear model
Published in: Journal of Public Health | Issue 5/2016
Login to get accessAbstract
Aim
The main purpose of this study is to introduce a high-accuracy model for predicting the best outcome of patients following liver transplantation.
Subject and methods
Computer-based medical prognosis is becoming increasingly significant as the volume of medical records increases every day, making manual processing harder. In addition, the inability of people to understand patterns from these huge volumes of data demands the use of machine learning tools. We propose an artificial neural network model to address the problem of organ allocation as well as survival prediction. This model extracts the relevant features and classifies the data set into training and test sets. Appropriate donor-recipient pairs were selected using ten-fold cross validation when training the medical data.
Results
An accuracy of 99.74 % was represented by a multilayer perceptron artificial neural network model. We could observe that the graft survival rate with our data set using the MELD score was 79.17 %. We also tested our model with three existing works containing different data sets and proved that the highest accuracy was obtained in the model with our data set.
Conclusion
To ensure accuracy, we made a comparison with existing models using various performance features. For training the model, we used a rich data set from the United Network for Organ Sharing transplant registry. We also carried out a survival analysis over 12 years, predicting survival probabilities using this rich data set and comparing it with existing approaches.