A. S. Krasichkov, A. A. Sokolova

Extrasystoles are premature impulses of ectopic origin and contraction of either the heart or its areas. The highest risk is associated with the ventricular extrasystoles (VE) that are generated by the impulses originating from the ventricular myocardium. In this paper, the statistical characteristics of electrocardiogram (ECG) in the presence of VEs are studied, and an algorithm to recognize this pathology based on the ECG analysis is suggested and experimentally tested. Here the ECG analysis aspects are twofold: the heart rhythm analysis and the QRS-complex duration analysis followed by a decision-making algorithm for a VE episode detection. In the rhythm analysis the main quantity is the quotient of the consecutive R-R intervals duration. The analysis is based on 39 recordings from an ECG database. The difference of distributions of this parameter in the presence and in the absence of VE allows to use this information in the decision-making procedure on the basis of rhythm parameter. To evaluate the QRS-complex features, two parameters have been used, the effective QRS-complex duration and its effective spectral bandwidth. The results of the comparison of the QRS-complex features in the time- and frequency- domains clearly indicate that the effecitve spectral bandwidth with the [0, 50] Hz band in the presence of miographic noise in superiorly robust to the signal shape variations under its constant duration. Therefore the algorithm can be synthesized on the basis of Neumann-Pearson criteria based on the comparison of the quotient of consecutive R-R in-tervals duration and QRS-complex parameters with relevant thresholding. Here we demand that the false alarm rate α is below a fixed threshold, and the threshold level can be varied such that for a fixed missed event rate β the minimum of the false alarm rate α is achieved. This approach allows not only to detect a VE, but on the basis of further analysis also to analyze the heart rhythm structure and to perform an extensive classification of VEs.