Prediction of Intrapartum Hypoxia from Cardiotocography Data Using Machine Learning

Uterine contractions produced during labor have the potential to damage a fetus by diminishing the maternal blood flow to the placenta, which can result in fetus hypoxia. In order to observe this phenomenon in practice, labor and delivery are routinely monitored using cardiotocography monitors. The cardiotocography recordings are used by obstetricians to help diagnosis fetus hypoxia. However, cardiotocography capture and interpretation is time consuming and subjective, often leading to misclassification that result in damage to the fetus and unnecessary cesarean sections. Therefore, correct classification is dependent on qualified and experienced obstetric and midwifery staff and their understanding of classification methods. As such, other, more objective measures may help to reduce misclassification among medical practitioners. For example, automatic detection of correlates between uterine contractions and fetal heart rate can be used to reduce unnecessary medical interventions, such as hypoxia and cesarean section, during the first stage of labor and can be instrumental in vaginal delivery in the second. The challenge is to develop predictive algorithms capable of detecting, with high accuracy, when a child is genuinely compromised before medical intervention is considered. This chapter aims to provide a discussion of the requirements for such an approach and discuss a methodology for objectively identifying intrapartum hypoxia.