Use of Neural Networks for Prediction of Graft Failure following Liver Transplantation

著者: Sherri Matis, Howard Doyle, Ignazio Marino, Richard J. Mural, Edward C. Uberbacher - Proceedings Eighth IEEE Symposium on Computer-Based Medical Systems 1995 被引用: 15

Liver transplantation k a well-established therapeutic option forpatients with end-stage liver dkease.However, up to 20% of transplanted livers fail to have adequate function initially, and at least harf of those will eventually fail.Accurate, early prediction of outcome may ameliorate thk situation by encouraging retransplantation before the patient's condition becomes irreversible.In thk shrdy, clinical information was gathered prospectively for 295 patients who underwent liver transplantation at the Universiy of Pittsburgh Medical Center, and was divided into sets.The feed-forward, fully connected, neural networks had 7 or 8 inputs, a single hidden layer conskting of 3 nodes and a single output node cfailure=I, success=O).The networks were trained with data from a randomly selected subset of 240 patients while the remaining 55 patients made up the test set.The preoperative (day 0) data conskted of patient demographics plus the results of standard liver function tests.The "day 1" data consisted information gathered during surgery plus the prediction of outcome from day 0. Data for days 2-5 included resultsfrom standard liver function tests plus the prediction of outcome from the previous day's network The network was trained using a standard back propagation algorithm.naining was assessed by testing the abiliy of the network to correctly predict the outcome of the 55 patients in the test set.The accuracy of prediction by the neural network improved each day and so by day 5, 98% of the grafi survivors in the test set were correctly predicted while 88% of grafi failures in the test set were correctly predicted.