论文部分内容阅读
基于自回归(AR)模型的心率变异性(HRV)分析技术广泛用于自主神经系统功能状态评价,其中AR模型阶数的选择对于HRV分析结果的准确性有重要的影响,本文研究了AR模型最佳阶数的确定方法。从46名健康成年受试者自然呼吸下的心电信号中提取心跳间期时间序列,用最终预测误差最小准则来计算AR模型最佳阶数,用Burg算法求解AR模型系数,并对残差序列进行白化检测来验证AR模型阶数的合理性。将该方法获得的HRV频域参数(包括总功率、低频功率、高频功率、低高频功率比以及标准化低频功率)与Kubios-HRV分析软件计算得到的结果作对照分析。结果表明通过该方法获得的HRV的5个频域参数均与Kubios-HRV分析软件的结果高度相关(相关系数r大于0.95),除总功率指标外,均无显著差异,对应的Bland-Altman图也有大于95%的点分布在一致性界限内。优化的基于AR模型的HRV分析算法能获得准确的HRV分析结果,与常用的HRV分析软件Kubios-HRV的结果有很好的一致性。
The heart rate variability (HRV) analysis technique based on autoregressive (AR) model is widely used in functional state evaluation of autonomic nervous system. The choice of AR model order has an important influence on the accuracy of HRV analysis results. In this paper, AR model The method of determining the best order. The time series of heartbeat interval was extracted from the ECG signals of 46 healthy adult subjects and the optimal order of AR model was calculated with the minimum final error prediction criterion. The AR model coefficients were solved by Burg algorithm and the residuals The sequence is whitened to verify the rationality of the AR model order. The HRV frequency domain parameters (including total power, low frequency power, high frequency power, low frequency power ratio and standardized low frequency power) obtained by this method were compared with the results calculated by Kubios-HRV analysis software. The results show that all the five frequency-domain parameters of HRV obtained by this method are highly correlated with the results of Kubios-HRV analysis software (the correlation coefficient r is greater than 0.95), with no significant difference except for the total power index. The corresponding Bland-Altman graph There are also more than 95% of the points distributed within the limits of consistency. The optimized HRV analysis algorithm based on the AR model can obtain accurate HRV analysis results, which is in good agreement with the results of the commonly used HRV analysis software Kubios-HRV.