Artificial neural networks using autoregressive parameters and periodogram samples were developed for ECG rhythm classification and compared with a learning vector quantization algorithm.
This paper discusses ECG rhythm classification using artificial neural networks (ANNs). We consider one 3-class problem where we distinguish between the normal sinus rhythm and two different abnormal rhythms, -and the practically very important "treat/no-treat" 2-class problem encountered e.g. when operating a semi-automatic defibrillation device. Autoregressive (AR) parameters, and samples of the signal's periodogram, are combined into feature vectors which are used as inputs to forward-connected multilayered perceptron ANNs. Training and testing is performed using signals from two different ECG data bases. The "best" net sizes and feature vector dimensions are decided upon by means of empirical tests. The results are compared with a previous method which also uses the AR model but which employs the learning vector quantization (LVQ) algorithm for the actual classification.
Øien et al. (Tue,) conducted a other in Arrhythmia. Artificial neural networks (multilayered perceptron) vs. Learning vector quantization (LVQ) algorithm was evaluated on ECG rhythm classification (3-class and 2-class problems). Artificial neural networks using autoregressive parameters and periodogram samples were developed for ECG rhythm classification and compared with a learning vector quantization algorithm.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: