Shariati M A, Dripps J H, Shariati H, Boddy K
University of Edinburgh, Department of Electical Engineering, UK.
Biomed Sci Instrum. 1993;29:161-8.
Long term fetal heart rate variability (LFHRV) is equivalent to detrended (when the time series trend removed) indirectly derived averaged fetal heart rate (FHR) time series. Up to now LFHRV is looked for visually in FHR traces, by obstetricians and midwifes for determination of fetal condition. The detrended averaged FHR data or LFHRV data is a random correlated non-stationary (in second moment) time series. In this paper we have applied linear stochastic time series analysis technique to identify a parsimonious stochastic model, for parametric numerical representation of the random cyclical patterns observed in short 2 minutes quasi-stationary contiguous blocks of LFHRV data. The parametric estimation technique used is based upon the optimum exact Maximum likelihood estimation (which uses Kalman filtering as part of its implementation). Diagnostics performed on the residuals indicated that a second order autoregressive model is a statistically adequate model in capturing variability patterns observed in 2 minute data windows of detrended average FHR. Also through further analysis of the spectral behavior of this identified model, pseudo-periodicity (or random periodicity) which for a long time was visually detected, can now be detected via this numerical procedure.
长期胎儿心率变异性（LFHRV）等同于去除趋势（当时间序列趋势被去除时）的间接推导的平均胎儿心率（FHR）时间序列。到目前为止，产科医生和助产士通过目视在FHR轨迹中寻找LFHRV，以确定胎儿状况。去除趋势的平均FHR数据或LFHRV数据是一个随机相关的非平稳（在二阶矩方面）时间序列。在本文中，我们应用线性随机时间序列分析技术来识别一个简约的随机模型，用于对在LFHRV数据的短2分钟准平稳连续块中观察到的随机周期性模式进行参数化数值表示。所使用的参数估计技术基于最优精确最大似然估计（其在实现过程中使用卡尔曼滤波）。对残差进行的诊断表明，二阶自回归模型在捕捉去除趋势的平均FHR的2分钟数据窗口中观察到的变异性模式方面是一个统计上合适的模型。此外，通过对这个识别出的模型的频谱行为的进一步分析，长期以来通过目视检测到的伪周期性（或随机周期性）现在可以通过这个数值程序检测到。
