ECG monitoring with artificial neural networks

For an electrocardiogram (ECG) monitoring system, it is desirable to have the capability of detecting heart abnormalities represented by waveshape changes within an ECG cycle in addition to detecting arrhythmias. The authors report their investigation of the use of artificial neural networks for ECG abnormality detection in an ECG monitoring scheme. Simulations are performed to explore how well neural networks can work after a short learning phase. The method used aims to produce a system which will perform well practically. It uses the waveform slope to locate the QRS complex for each ECG cycle which allows a small training data set to be formed for each individual patient. The performance of three neural network models is compared and results of the simulation show that the correct recognition of normal cycles and VPB cycles is typically greater than 95%