Time warping of apneic ECG signals using genetic algorithm

We construct a method of time warping in quasiperiodic time series analysis using genetic algorithm in order to extract the instantaneous phase difference between a template signal and a testing signal. Contrast to previous studies, which involves correlation estimations to determine the shape similarity of two signals taken from the quasiperiodic time series, time warping perform the comparison of the two signals by first constructing a discrete set of M points formed from uniformly sampled values of the template signal f(t). The discrete set of sample values of the testing signal, g(t'), which contains N points, will be interpolated to form a continuous function so that the difference between the template signal at those M points and the corresponding testing signals are minimize to best preserve the mapping of the two signals. The result of this optimization procedure produces a phase shift function that relates the time t' in the testing signal to the time t in the template signal. Due to the numerous choices in the partitioning of the time domain of the two signals, genetic algorithm is found to be effective in extracting this phase shift function. We apply this theoretical tool of time warping using genetic algorithm to analyze the electrocardiographic (ECG) signals, with the aim to investigate if central apneic and obstructive apneic episodes can be differentiated from non-apneic episodes. Detailed statistical analysis of the phase shift from real ECG data of sleep apnea patient indicates that the difference of both magnitude and phase of the signals can be used to differentiate apneic events from non-apneic events.