Classification of cardiac arrhythmias based on principal component analysis and feedforward neural networks

In previous work (see Proc. 13^th Ann. Int. Conf. of IEEE/EMBS, vol. 1, p. 580-1, 1991), the authors created a database containing the first 10 principal component coefficients and the relative RR intervals of P-QRS complexes from all the patients of the MIT-BIH Arrhythmia Database. Here, the authors use logistic regression and feedforward neural networks for classifying the heart beats of patient 208, based on these principal component coefficients. The feedforward neural network technique is presented as an extension of the concept of logistic regression, applicable for classifying patterns into more than two classes. The results indicate the potential of the approach for the classification of cardiac arrhythmias