Fundamental Analyses of Ventricular Fibrillation Signals by Parametric, Nonparametric, and Dynamical Methods

Abstract

Ventricular fibrillation (VF) is the malignant electrical rhythm of the heart. Fundamental understanding of this rhythm can only be obtained by considering signals generated by single heart cells, isolated heart tissue, and the whole organ. Interpretation of these complex phenomena also requires that we employ the modern signal processing methods that consider the temporal, spectral and dynamical features of this rhythm. We recorded action potentials (AP) from single cells in isolated fibrillating hearts with the aid of a floating microelectrode technique, and in cardiac tissue using optical fluorescence imaging. 1) Time-frequency analysis of these signals reveals different characteristics of VF signal during early and late stages of fibrillation. Perfusion of the heart maximized the short-term, high-frequency events, while without perfusion, the time-frequency distributions showed dispersion. Time-frequency analysis was thus shown to be helpful in characterizing AP during various stages of VF. 2) Parametric modeling was next considered to determine the evolution of fibrillation with extended periods of time, a situation that would arise during resuscitation procedures. Autoregressive modeling of VF signals was carried out to identify changes in the dominant poles of the VF signals with time course of evolution of VF. Parametric modeling of VF was thus shown to be useful in predicting the duration of cardiac arrest. 3) Finally, we sought to determine how dynamics of VF signals change with time, and in particular whether fibrillation can be considered chaotic at the cellular levels. Dynamical analysis, carried out by the methods of dimensional analysis and Lyapunov exponents revealed that VF has a relatively low dimensional attractor at the single cell level even though VF on the heart or the body surface may be a high dimensional process. Such analyses may help suggest methods to track and modify low dimensional chaotic VF in its early stages of evolution. 4) Finally, algorithms were developed for application in a clinical device such as the implantable cardioverter-defibrillator. Here, the emphasis is on the reducing false positive and false negative rates. This objective is accomplished using a sequential hypothesis testing algorithm that trades off accuracy for detection time and vice versa. In summary, non-parametric, parametric, and dynamical methods together provide quantitative insights into the fibrillation phenomenon at the cellular and whole heart levels, and may help in discrimination of various stages and forms of VF for the purposes of possible therapy.