Automatic diagnosis of heart diseases using neural network

Abstract

The use of artificial neural networks for classification of anteroseptal myocardial infarction (ASMI) from the electrocardiogram (ECG) is investigated. The ECGs of ASMI cases and nonASMI cases including normals have been collected and are represented by 'complete trees'. ECG morphology features have been extracted from the individual tree for classification. A three layer back-propagation trained neural network, based on a gradient descent algorithm is used for classification of ASMI cases from others. The network has been trained with features extracted from the V/sub 1/, V/sub 2/ and V/sub 3/ ECG leads of thirty cases of known ASMI and thirty cases of nonASMI. The performance of the network was evaluated by comparing the results obtained from the network with clinical results.