Comparison of the effect of different sampling modes on computer analysis of cardiotocograms

Cardiotocographic (CTG) monitors may provide fetal heart rate (FHR) signals as beat-to-beat (BTB) or alternatively at a fixed sampling rate. The aim of this study was to assess the effect of different sampling modes on the evaluation provided by a commercially available system for computer analysis of CTGs.

Internal FHR signals were acquired during the last hour of labor in 27 singleton term cephalic pregnancies, using the STAN S31(®) fetal monitor (Neoventa, Gothemburg, Sweden). BTB and 4 Hz sampling outputs of the monitor were compared using the Omniview-SisPorto(®) system for computer analysis of CTGs (Speculum, Lisbon, Portugal). The following parameters were analyzed: signal loss, signal quality, baseline, accelerations, decelerations, percentage of abnormal short-term variability (%aSTV), abnormal long-term variability (%aLTV), average short-term variability (avSTV) and the system's clinical alerts. Statistical inference was performed using the Spearman correlation coefficient, 95% nonparametric confidence intervals, Wilcoxon and McNemar statistical tests, setting significance at 0.05, and a non-parametric measure of disagreement (valued 0-1 from lowest to highest disagreement).

Comparing BTB with 4 Hz sampling, the median values for signal quality (95% versus 96%), number of accelerations (5 versus 7) and %aSTV (31 versus 39) were significantly lower in the former. On the other hand, with BTB signals the median value of avSTV was significantly higher (3.1 versus 2.3). Nevertheless, BTB and 4 Hz parameters were highly correlated (r=0.89-0.97), and there were no significant differences in the quantification of the number of decelerations or in the clinical alerts elicited by the system.

In conclusion, different sampling modes have a significant effect on the parameters provided by computer analysis of CTGs that are related with the quantification of STV, with a small impact on baseline estimation and on the subsequent quantification of accelerations. However, there does not seem to be significant impact on the quantification of decelerations or on the alerts provided by the system.