DIAGNOSTIC ECG CLASSIFICATION BASED ON NEURAL NETWORKS

This study illustrates the use of the neural network approach in the p roblem of diagnostic classification of resting 12-lead electrocardiogr ams. A large electrocardiographic library (the CORDA database establis hed at the University of Leuven, Belgium) has been utilized in this st udy, whose classification is validated by electrocardiographic-indepen dent clinical data. In particular, a subset of 3,253 electrocardiograp hic signals with single diseases has been selected. Seven diagnostic c lasses have been considered: normal, left, right, and biventricular hy pertrophy, and anterior, inferior, and combined myocardial infarction. The basic architecture used is a feed-forward neural network and the backpropagation algorithm for the training phase. Sensitivity, specifi city, total accuracy, and partial accuracy are the indices used for te sting and comparing the results with classical methodologies. In order to validate this approach, the accuracy of two statistical models (li near discriminant analysis and logistic discriminant analysis) tune on the same dataset have been taken as the reference piont. Several nets have been trained, either adjusting some components of the architectu re of the networks, considering subsets and clusters of the original l earning set, or combining different neural networks. The results have confirmed the potentially and good performance of the connectionist ap proach when compared with classical methodologies.