Multi-component based neural network beat detection in electrocardiogram analysis

Electrocardiogram (ECG) classification systems have the potential to benefit from the inclusion of automated measurement capabilities. The first stage in the computerized processing of the ECG is Beat Detection. The accuracy of the beat detector is very important for the overall system performance hence there is benefit in improving its accuracy. In the present study we introduce the concept of a multi-component based approach to beat detection based on neural networks (NNs). A database containing in excess of approximately 3000 cardiac cycles was used to evaluate the techniques developed. Results showed the enhanced capability of the multi- component based approaches to detect up to 2988 beats in comparison to 2848 beats achieved by standard benchmarking techniques of non-syntactic and cross- correlation methods. These results have subsequently demonstrated the improvements which can be achieved through utilization of the proposed approach.