Estimating respiratory frequency from autoregressive spectral analysis of heart period.

The noninvasive assessment of intrinsic sources of heart period (HP) variability is an area of great interest in medicine and physiology [1]. Spectral analysis of HP variability is a method that has been widely recommended and employed in deriving indices that reflect sympathetic and parasympathetic cardiovascular influences [2]-[5]. This technique extracts from beat-to-beat variations in HP spectral components that differentially reflect autonomic mediators of cardiovascular variability. A fast, high-frequency (HF) component (−0.25 Hz) corresponding to the frequency of respiration (reflecting cardiac vagal activity) and a slower, low-frequency (LF) component (−0.10 Hz) due primarily to baroreceptor-mediated regulation of blood pressure have been consistently found in the power spectrum of the electrocardiogram (ECG) [4], [6]. LF spectral power has been speculated to primarily reflect sympathetic activity [7], [9], though varying levels of parasympathetic influence can been found at this frequency under certain conditions [5], [10]. The central frequencies of these spectral components have also been suggested as useful indices of physiological function. Recently, the central frequency of the LF component has been found to be decreased in patients with autonomic dysfunction [l]. Importantly, the central frequency of the HF component may serve as an index of respiratory frequency (RF) when more direct measures are not available. Given that some researchers have suggested that measures of HP variability may need to be corrected for respiratory parameters such as RF [11], the utility of this index needs to be investigated. This may be particularly useful in ambulatory settings where the strain gauge measurement of respiration might be difficult or compromised. In this article we report the results of a pilot study and a larger investigation that examined the relationship between respiration frequency assessed using the traditional mercury strain gauge and using the central frequency of the HF component derived from autoregressive spectral analysis.