Myocardial infarction: diagnosis and vital status prediction using neural networks

Abstract

Neural networks (NNs) have been found useful in many biomedical applications. The authors' purpose is to apply NNs to two specific problems in cardiology, namely, diagnosis of echocardiograms for myocardial infarction and prediction of vital status of patients that suffered such. The authors used NNs to discriminate between normal and infarcted myocardium, by looking at intensity changes. The intensities of selected regions are used for training and testing. In predicting the vital status of patients that have suffered acute myocardial infarction, the authors used a large database (MIDAS) with follow-ups. The NN in this case has two hidden layers with 18 patient variables from the MIDAS dataset as inputs. The NN was again trained with the feedback algorithm ALOPEX and tested with unknown data.<>