| Summary: | 碩士 === 國立陽明大學 === 醫學工程研究所 === 93 === Atrial flutter (AFL, including typical and atypical) and atrial fibrillation (Af) have different mechanisms in atrium. It is difficult to distinguish them from the surface ECG. Nonlinear analysis has recently been applied to cardiac signal analysis. However, the assumption that the atrial electrical conduction corresponds to a nonlinear dynamic system is still controversial. In this study, we applied surrogate data testing to analyze the properties of electrocardiograms with atrial flutter and fibrillation in humans. The systems were divided into two types: nonlinear dynamic system and linear Gaussian stochastic system.
Methods:
By using independent component analysis, we obtained atrial activities from 12-lead surface electrocardiograms. Twenty AFLs (including 10 typical and 10 atypical) and 7 Afs were analyzed. The ‘iterative amplitude-adjusted Fourier transform algorithm’ was applied to generate surrogate data. Time-reversibility (trev), correlation dimension (D2), and maximal Lyapunov exponent (λ1) were calculated as discriminating parameters. Finally, accepting or rejecting null hypothesis was decided by rank-order test. The property of atrial systems was explained by the result of surrogate data testing. Nonlinearity was emphasized by time-reversibility. Complexity of a system was represented by correlation dimension. The sensibility to initial condition was quantified by maximal Lyapunov exponent.
Results:
In these three types of arrhythmias, for typical flutters, the properties of nonlinear dynamics were significant. When we used trev, D2 and λ1 to test, the proportions of null hypothesis rejected were 7/10, 8/10 and 10/10, respectively. For Af, ECG patterns became complicated. When we used trev, D2 and λ1 to test, the proportions of null hypotheses rejected were 1/7, 0/7 and 0/7, respectively. During atypical flutter, ECG showed more regular than Af but more complex than typical flutter. Thus, when we used trev, D2 and λ1 to test, the proportions of null hypotheses rejected was 4/10, 8/10 and 4/10, respectively.
Conclusion:
The systems of typical flutter, atypical flutter and Af were distinct. By using the surrogate data testing, we found that typical flutter was a nonlinear dynamic system, and Af was a linear Gaussian stochastic system.
|
|---|
