Proposal of sudden death indicators based on information measures from ECG signals

Sudden death (SD) of millions of people every year is a worldwide problem of interest for researchers and scientists. There is a lack of evidence in the specific literature about the waveform that precedes the event of SD. On the other hand, it can be found a large amount of publications carried out in the time-domain, such as Hearth Rate Variability (HRV) or interlatency analysis. The aim of this work is to present a pilot study that may provide important clues to the characterize the ECG signal of the SD syndrome. Given a database containing electrocardiograms (ECG) records from people either with normal sinus rhythm (NS) or who died suddenly, for each windowed signal the normalized Bandt and Pompe entropy is calculated using an embedding time delay that depends on the signal. Then, the behavior of the autocorrelation of the entropy of both groups of signals is studied showing a promising way to differentiate them.