Abstract:
The cardiac activity can be investigated based on the RR signal - a series of temporal intervals between consecutive heartbeats. The variation of these intervals - called the heart rate variability (HRV) - enables quantitative analysis of functioning of the cardiac control mechanism (Autonomous Nervous System). The mainstream techniques of the HRV analysis are time-consuming and do not identify conditions possibly affecting the HRV, what limits their analytical scope and depth. Hence the demand from cardiological community to develop reliable methods of HRV assessment working in nearly-real time and based on relatively small amount of data, yet at the same time providing nontrivial insight into the cardiac activity. We address this problem by investigating the dynamical changes in nonstationary RR signal, extracted from electrocardiogram recorded in presence of controlled environmental stimuli, including music. For this purpose we introduced the Sample Entropy-related methods to quantify complexity in time series, which we applied independently as (i) change detectors and (ii) classification features. Furthermore, we propose and demonstrate methods of symbolic analysis of the RR signals based on the notion of Lempel-Ziv complexity. Our research has laid the foundation for using novel nonlinear-dynamical statistics implemented as change detection algorithms applicable to the HRV analysis. We have shown that the new methods are sensitive enough to capture effects of subtle stimuli - such as music - on the HRV characteristics, while not compromising robustness to noise and experimental artefacts. Such techniques can find application in a variety of domains beyond cardiology, where the identification of change can be a starting point for the stress detection. In a wider perspective, the methods are potentially applicable to detecting nonstationarity in the systems whose dynamical parameters drift over time.
