A Multiparameter Statistical Approach for Cardiotocographic Signals Analysis

Cardiotocography (CTG) is a widely way to assess foetal well-being status, both in antepartum and intrapartum period. Automatic computerized analysis was introduced to overcome limitations of visual analysis of CTG traces. The present study extended a software previously developed for the calculation of foetal heart rate signal (FHR), floatingline (FL) signal and uterine contractions signal introducing the extraction of five novel statistical parameters from signals mentioned. Dataset contained 580 signals acquired during daily routing foetal monitoring. A univariate statistical analysis was used to compare spontaneous deliveries from caesarean deliveries; an analysis of variance, and a post-hoc test was performed to distinguish the week of gestation (WG) in which the CTG trace was recorded. The results demonstrated that, whereas decelerations features were not able to classify CTG signals, accelerations features were. In particular, kurtosis of FHR accelerations (p-value=0.0029) was able to distinguish between CTG corresponding to caesarean and spontaneous delivery. Furthermore, by considering accelerations computed by the FL, their mean (p-value=0.034), standard deviation (p-value=0.032) and variance (p-value=0.011) were useful in discriminating among different WG classes. These are important and promising results because it is the first time that novel features are employed in CTG classification.