Abstract:
Neural Networks, a sub-field of Artificial Intelligence, has been widely used in classification with some success. Accurate analysis is an important process in classification of medical data. This Special Project Study is concerned with developing a computer system to classify different stages of heart condition. Three types of Feedforward neural networks, namely Multi-layer Perceptron, Generalized Feedforward and Modular networks, had been employed in the classification. Back Propagation learning algorithm was adopted in all three neural network structures. Training and testing data sets were readings from monitoring electrocardiograms. These readings were collected from four different hospitals in Thailand. Neural Networks developed are capable of classifying 4 different stages of heart condition (1 normality and 3 abnormalities). A neural network software known as NeuroSolutions was used as the tool for development. The result reveals that Multi-layer Perceptron achieved the highest level of accuracy (93.93%). This Special Project Study confirms the potential Feedforward neural networks and suggests another alternative to application of neural networks technology in electrocardiography analysis.