Statistical Modeling of Fetal Heart Rate Variability

The random variability of fetal heart rate (FHR) is of practical importance in obstetric management. The variability is not white, however, and cannot be adequately described by its variance alone. We have modeled FHR variability as the scalar output of a multivariate, nonlinear stochastic process. An adaptation of the Kalman filter algorithm is used as part of an approximate maximum likelihood estimator of the process parameters. The filter algorithm permits recording artifacts to be disregarded, by treating them as "missing" observations.