Improving the fetal cardiotocographic monitoring by advanced signal processing

The paper aims at predicting possible pathologies at birth on the basis of fetal heart rate variability analysis performed on the cardiotocographic recordings in the antepartum period. Data belong to a database including more than 800 tracings. Fetal condition is usually classified by clinicians as normal or suffering, when signs of fetal distress are clearly recognized. Nevertheless some suffering fetuses remain unclassified due to the weakness of the clinical tools, actually supporting the analysis. Fetal distress can be an early marker of pathology but it constitutes per se a risk alert for the fetal monitoring. In this work we propose new analysis tools, based on heart rate variability signal processing, which try to identify some markers leading the fetal sufferance to evolve into a pathological condition. Parameters in the time domain as well as nonlinear estimators such as approximate and multiscale entropies have been employed. Results evidence significant differences among the normal and suffering/pathological groups. Clustering the data by the gestational week and the Apgar score can further improve the analysis.