Sapoznikov D, Luria M H, Gotsman M S
Department of Cardiology, Hadassah University Hospital, Jerusalem, Israel.
Comput Biomed Res. 1994 Jun;27(3):199-209. doi: 10.1006/cbmr.1994.1017.
Periodic low-frequency (LF) fluctuations of heart rate (HR) may be of diagnostic and prognostic value in diverse pathologic cardiopulmonary conditions. Two principal components of LF HR rate variations may be distinguished: periodic fluctuations and nonperiodic, nonstationary changes. The frequency content of these two components may overlap considerably. In order to avoid a tedious work-intensive visual analysis an efficient computer-based method for detection, differentiation, and quantitation of these signals is required. Two methods for separating periodic from nonperiodic HR changes are presented, namely, detrending and bandwidth (BW) calculation. A group of healthy individuals was evaluated in order to assess these methods in individuals with significant LF periodic episodes (15 patients) contrasted to those without LF periodic episodes (94 patients). The commonly used method of detrending consists of a fitted polynomial which by subtraction removes low frequencies originating from nonstationary changes without affecting periodic fluctuations. We found, however, that the frequencies involved in nonstationary and periodic fluctuations often overlap and thus the detrending method may not be highly efficient. In a second method we postulated different shapes for power spectrum curves of periodic and nonperiodic episodes. This latter method is based on BW calculation of the LF component of the R-R power spectrum and proved to be more efficient in detecting periodic episodes. It showed higher significance levels for the difference between the periodic and nonperiodic groups when the BW or the ratio between peak power and BW in the LF range was used. This new, alternative detection method may be employed in further studies which seek to elucidate the clinical relevance of the LF range and, in particular, the mechanisms for such long-wavelength periodic fluctuations.
心率（HR）的周期性低频（LF）波动在多种病理性心肺疾病中可能具有诊断和预后价值。LF心率变化的两个主要成分可以区分：周期性波动和非周期性、非平稳变化。这两个成分的频率内容可能有相当大的重叠。为了避免繁琐的高强度视觉分析，需要一种基于计算机的有效方法来检测、区分和量化这些信号。本文介绍了两种将周期性HR变化与非周期性HR变化分离的方法，即去趋势分析和带宽（BW）计算。为了评估这些方法，对一组健康个体进行了评估，将有明显LF周期性发作的个体（15例患者）与无LF周期性发作的个体（94例患者）进行对比。常用的去趋势分析方法包括拟合多项式，通过减法去除源自非平稳变化的低频，而不影响周期性波动。然而，我们发现，非平稳波动和周期性波动所涉及的频率常常重叠，因此去趋势分析方法可能效率不高。在第二种方法中，我们假设周期性和非周期性发作的功率谱曲线具有不同的形状。后一种方法基于R-R功率谱LF成分的BW计算，在检测周期性发作方面被证明更有效。当使用BW或LF范围内的峰值功率与BW之比时，它显示出周期性组和非周期性组之间差异的更高显著性水平。这种新的替代检测方法可用于进一步的研究，以阐明LF范围的临床相关性，特别是这种长波长周期性波动的机制。
