Application of neural networks to postoperative liver transplant monitoring

This paper explores the feasibility and efficacy of applying artificial neural network technology to assist with the clinical management of human liver transplant recipients. We describe a novel application of neural network technology to this domain and present results from three experiments which assess the performance gains obtained. These experiments directly compare the statistical techniques, logistic regression and discriminant analysis with multi-layer perceptrons (MLPs) for performing rejection risk assessment. This paper documents an analysis of progressively more sophisticated modelling techniques, together with a discussion of the advantages and disadvantages of each approach. These experiments lead us to conclude that MLPs offer significant advantages over traditional statistical methods in this domain. Finally, the paper introduces a discussion, together with interim results, of the future directions being explored in this research program. In particular, this includes the use of temporal information to further enhance the performance of the most promising of the connectionist systems described here.