心率变异性谱分析方法的比较研究

A Comparison of Periodogram, BT and AR Techniques for Spectral Analysis of Heart Rate Variability

为了分析比较不同谱估计方法及数据长度对心率变异性（ＨＲＶ）信号谱估计结果的影响，对健康被试者不同体位条件下获取的ＨＲＶ信号，选取４种不同的数据长度，分别用周期图、自相关（ＢＴ）及自回归参数模型（ＡＲ）法进行谱估计，并对所得结果进行了统计分析。结果表明：（ｌ）在相同体位条件下，相同谱估计方法选取４种不同数据长度估计得到的谱参数之间无显著性差异；（２）在相同数据长度，相同体位条件下，分别采用周期图与ＡＲ方法得到的谱参数之间无显著性差异，与ＢＴ法比较，则存在显著差异；（３）由平卧转为站立，反映迷走神经活动水平的高频谱峰功率显著降低，低频谱峰功率变化不显著，谱的总功率显著降低，低、高频谱峰功率比则升高。

Heart rate variability (HRV)reflects both the presence of certain va-riety of naturally occurring physiological perturbations to cardiovascular homeostasis andthe dynamic response of the cardiovascular control systems to these perturbations . Spec-tral estimation techniques have recently been used to study the HRV signals. It has beenshown that the HRV power spectrum provides a noninvasive tool to monitor the autonom-ic control on the heart both in normal and pathological subjects. To elucidate the effectsof different spectral estimation techniques and the data length on the HRV spectral analy-sis , three spectral estimation methods with four data lengths were applied to HRV signalsobtained from ten healthy young subjects in both supine and standing positions, The threetechniques were(1)the periodogram method,(2 )the Blackman-Tukey(BT) method,and(3 )the autoregression(AR ) method.AIC was used to determine the appropriatemodel order. The four data lengths of the HRV signal were 64, 128 ,256 ,and 360points.Based on results from the literature and our preliminary studies,four spectral in-dexes were defined and calculated which included a low frequency component(LF),ahigh frequency component(HF),a total power component(TP) ,and a LF to HF ratio(RF=LF: HF).Experiments and spectral estimation results indicated that(1) the spec-tral indexes estimated with the same method and in the same body position did not showany significant differences among the four data lengths,(2) the spectral indexes estimat-ed with both the periodogram and AR methods in the same body position and for the samedata length were statistically identical. But , when compared with the BT method thereexisted significant differences ,and(3 ) The HF and TP decreased and the RF increasedsignificantly(P<0.05 ) when the subjects moved from supine fo standing position. Fromthese findings we conclude that(1) short data length may be suitable for the HRV spec-tral analysis,(2 ) the spectral indexes estimated with either periodogram or AR methodare comparable when the original data were obtained from different physiological experi-ments, and(3 ) the autonomic nervous system activity associated with changes of bodyposition can be quantitatively assessed by the HRV spectral analysis.