Prediction and prognosis of acute myocardial infarction in patients with previous coronary artery bypass grafting using neural network model

Abstract

Introduction/Objective. The aim of this study was to analyze the usefulness and accuracy of artificial neural networks in the prognosis of infarcted patients with previous myocardial surgical revascularization. Methods. The 13 predictor variables per patient were defined as a data set. All the patients were divided into two groups randomly: the training group and the test group, of 1090 patients each. The evaluation of the neural network performance was organized by using the original data, as well as the complementary test data, containing patient data not used for training the network. In generating the file of comparative results, the program compared the actual outcome for each patient with the predicted one. Results. All the results were compared with 2 ? 2 contingency table constructed from sensitivity, specificity, accuracy, and positive?negative prediction. The network was able to predict the outcome with the accuracy of 96.2%, sensitivity of 78.4%, specificity of 100%, positive predictivity of 100%, and negative predictivity of 96%. There was not efficient prognosis of infarcted patients previously operated on using linear discriminant analysis (accuracy 68.3%, sensitivity 66.4%, and positive predictivity 30.2%). Conclusion. This study suggest that a neural network was better for almost all parameters in outcome prognosis of infarcted patients with previous myocardial surgical revascularization.