Classification of fetal and neonatal heart rate patterns in relation to behavioural states

Abstract

For the assessment of behavioural states in the human fetus, the fetal heart rate (FHR) pattern is one of the state variables. A statistical method is described to classify FHR patterns. FHR recordings were made between 38 and 40 wk gestation. The tachogram was averaged over 3-s intervals. For FHR segments of 3 min duration the parameters of an autoregressive-moving average (ARMA) model were estimated. Simulated FHR patterns, generated by using these estimated ARMA parameters, resembled real recordings. The ARMA parameters were used as features for a retrospective classification of the FHR segments, using a linear discriminant function. The classification by the above method was compared with an independent visual classification of the FHR patterns. The computer/observer classification agreement was 85% (kappa = 0.70). These data were compared with classification results for neonatal heart rate segments. For prospective classification of FHR patterns a moving discriminant function was introduced.