Bimodality and scaling in recurrence networks from ECG data

Abstract

Human heart is a complex system that can be studied using its electrical activity recorded as Electrocardiogram (ECG). Any variations or anomalies in the ECG can indicate abnormalities in the cardiac dynamics. In this work, we present a detailed analysis of ECG data using the framework of recurrence network (RN). We show how the measures of the recurrence networks constructed from ECG data sets, can quantify the complexity and variability underlying the data. Our study shows for the first time that the RN from ECG show the unique feature of bimodality in their degree distribution. We relate this to the complex dynamics underlying the cardiac system, with structures at two spatial scales. We also show that that there is relevant information to be extracted from the scaling of measures with recurrence threshold ε. Thus we observe two scaling regions in the link density for ECG data which is compared with scaling in RNs from standard chaotic and hyperchaotic systems and noise. While both bimodality and scaling are common features of RNs from all types of ECG data, we find disease specific variations in them can be quantified.

Multiplex Recurrence Networks from multi-lead ECG data

“…Then a recurrence threshold (ε) is chosen to construct the links between nodes and to arrive at the adjacency matrix of the recurrence network 27,28 . The variation of network measures with recurrence threshold itself is an indicator of the complexity of the underlying attractor, as reported recently 13 . In the present work we set it to ε = 0.1 for the sake of uniformity, based on previous study.…”

“…It transforms the recurrence pattern in the reconstructed dynamics from a given time series into a complex network 11,12 . In this context we have recently reported bimodality in the degree distribution and scaling of link density with recurrence threshold as characteristic features of RNs from ECG 13 .…”

Detection of healthy and pathological heartbeat dynamics in ECG signals using multivariate recurrence networks with multiple scale factors

“…The recurrence plot is a versatile and reliable method that has been extensively used to graphically represent the recurrence dynamics in time series data, which characterizes the proximity of two states in the time-delay reconstructed state space of a dynamical system. Because the recurrence of trajectories in phase space is one of the most important fundamental features of dynamical systems, consequently, the recurrence network method has been successfully applied to evaluate the dynamic behavior of real systems, such as climate, [19,20] the cardiac system, [21,22] electroencephalogram signals, [23,24] etc. However, when the number of time series is greater than 2, the relationship between the time series cannot be portrayed.…”