The use of the nonlinear features in arrhythmias detection with short heart beat time series
TIAN ZHAOYIN ( 1651215 )
Classic signal processing methods, like adaptive filtering, sequential hypothesis testing and ECG morphological features applying, can only provide limited information, because they ignore the nonlinear dynamic characteristics of the signal. The physiological signals used to be of non-stationary nature and the complex regulation can be expressed by the nonlinearity features, like the R-R interval, heart rate variety. Therefore, the nonlinear parameters can better reveal their characteristics and mechanism. Attempts to use the nonlinear parameters to analysis arrhythmias have been tried. Inspired by these researches, we tried to use the HRV-related parameters to capture distinct characteristics of different arrhythmias. We firstly tried to use Symbolic Coding and Shannon Entropy to distinguish the atrial fibrillation (AF) from normal sinus rhythm (N) ECG signal, this algorithm could easily separate AF from N with a high accuracy which can reach about 95%. After the ectopic heartbeats (V) ECG signals being added into the classification task, this algorithm didn't give clear separation for the 3 types of heartbeat. Specifically, the distribution of Shannon Entropy value of V consistently overlapped with both N and AF, which meant that this parameter can not distinguish the ventricular ectopic heartbeats (V) from atrial fibrillation (AF) and normal sinus rhythm (N). Secondly, we used the Multiscale Entropy (MultiEn) to try to get a more general description. By using the Refined Composite Multiscale Entropy, which is a modification of MultiEn, the distinct properties of the three arrhythmias (N, AF and V) can be captured. They showed different trend from scale 1 to scale 20. The trend of AF was that the value of Refined Composite Multiscale Entropy went down continuously. The trends of V and N were that the value of Refined Composite Multiscale Entropy would decrease at lower scales and turn to stable at higher scales. The difference between V and N was that the value of Refined Composite Multiscale Entropy of N was bigger that of V. At last, we applied this algorithm into shorter heart rate sequence and almost got the same results. Hence, we consider that this algorithm has significant meaning in real time arrhythmia detecting and could probably serve as an independent parameter for arrhythmia diagnosis.