AI Framework for Fetal Health Risk Prediction

Cardiotocography is an important technique used in the field of obstetrics to monitor the fetal heart rate and the mother’s uterine contractions throughout a pregnancy. Analysis of cardiotocographs allows for thorough understanding of fetal health, and thus, can improve fetal and maternal health outcomes. Traditional methods of monitoring fetal movement and uterine contractions through human analysis of cardiotocographs is not always reliable since this analysis is often subjective. Using Artificial Intelligence to analyze this data would, for these reasons, allow for more accurate predictions about the health of the fetus. Data consisting of 2126 cardiotocograph recordings of pregnant women has been taken from the University of California, Irvine and the University of Porto to conduct a study in monitoring fetal health using Artificial Intelligence. This study uses both IBM Watson and Python-based AI models to predict fetal health. The primary outcome of this study is a model that is trained to predict fetal health outcomes. The model provides 3 predictions: the number 1 corresponds to a predicted healthy fetus; a prediction of 2 corresponds to a fetus that suspectedly has a health problem, but it is not known for sure; the number 3 corresponds to a definitive health problem for the fetus. This study has found that the CatBoostClassifier coded in Python has a 95.3% accuracy, the highest of all models used.