Application Of Nonlinear Dynamics In AF Research

Atrial fibrillation (AF), the most common cardiac rhythm disturbance, is increasing in prevalence as the population ages. Automatically analyzing the ECG signals and diagnosing AF with computer is a significant job.A kind of binarization algorithm is provided in the paper, which can remove background noise and background grid, while keep the useful ECG signals. Firstly the indexed images in database are converted to RGB ones, then the corresponding noise is removed in R-array, G-array and B-array separately. Finally OR operation is experimented with the signals in three gotten arrays to get the result ECG. The paper develops a kind of algorithm to sample the signals saved in images so as to get the ECG signals. A filter with high-frequency disturbance depressed and a filter with baseline drifting depressed are designed with the help of Sptool tool box in MatLAB. High-frequency disturbance and baseline drifting can be efficiently depressed by the filters.Fractal dimension is inducted by the author to analyze AF ECG signals. Fractal dimension of a few AF ECG images and normal ones is calculated. The results show that there is difference between the average of fractal dimension of two kinds of images , but there is intersection among the fractal dimension of individual images.Identification of R wave is a decisive step in ECG signals' automatical analysis. A kind of algorithm is provided in the paper to identify the R wave, which combines forward-backward difference value algorithm, wave width algorithm and R-R interval algorithm. After experiment, the minimum identification exactness of AF ECG signals is 97.7%, the maximum is 100%, the average is 99.5%, the minimum of normal ECG signals is 99.7%, the maximum is 100%, the average is 99.9%. The algorithm is in possession of the ability of anti-disturbance and high exactness.Poincare plot is introduced at the end of the paper. Eleven Poincare plots of AFECG and thirteen Poincare plots of normal ECG are drawn using the MatLAB language. The results show that shapes of two kinds of plots have difference: Plots of AF ECG have short bar shape, fan shape and complex shape. Plots of paroxysmal AF are different from those of persistent AF. Most plots of normal ECG have comet shape, some present torpedo shape and few have wide-waist shape. LA vs. SA in Poincare plots is calculated in the paper, and the the conclusion is drawn that LA vs. SA in normal Poincare plots is different from that in AF ones. LA vs. SA can be used to distinguish AF ECG images from normal ones.