Summary
Objectives:
Heart-rate variability (HRV) is an interesting tool for assessing cardiac autonomic system control, but nonstationarities raise problematic issues. The objective of this paper is to show that adapted signal processing tools may cope with nonstationary situations and improve the analysis of HRV.
Methods:
We propose to use the recent method of Empirical Mode Decomposition (EMD), so as to analyze the cardiac sympatho-vagal balance on automatically extracted modes. The method, which is fully data-adaptive, consists in an iterative decomposition based on the idea that any signal can be locally represented as an oscillation superimposed to a more regular trend. When a signal is composed of distinct nonstationary components, EMD therefore achieves a time-varying filtering which effectively separates them.
Results:
The method has been applied to situations where postural changes occur, provoking instantaneous changes in heart rate as a result of autonomic modifications. In the considered application where the sympatho-vagal balance is quantified by comparing the low-frequency (LF) and high-frequency (HF) components of RR intervals, EMD automatically achieves a separation of these components upon which further processing can be carried. Visualizing the decomposition in the time-frequency plane, we can identify local events due to the postural changes, and we can assess a (time-varying) HF vs. LF discrimination without resorting to some fixed high-pass/low-pass filtering.
Conclusion:
Assessing cardiovascular autonomic control by resorting to LF/HF measurements may prove difficult in nonstationary situations where the use of a priori fixed filters can be questioned. Because it is both local and fully data-adaptive, EMD appears as an appealing and versatile pre-processing technique for overcoming some of the limitations that conventional spectral methods are faced with in nonstationary situations.
Keywords
Empirical mode decomposition - heart rate variability - sympatho-vagal balance
