Abstract:
This study investigates the application of machine learning techniques to predict
fetal health using Cardiotocogram (CTG) data, a critical tool in obstetrics for
monitoring fetal well-being. CTG captures vital indicators such as fetal heart
rate (FHR), uterine contractions, and fetal movements, which are essential for
assessing the fetus’s condition during pregnancy and labor. Early detection of
potential fetal distress is crucial in preventing complications and reducing
maternal and neonatal mortality rates. However, the manual interpretation of
CTG data can be time-consuming and error-prone, highlighting the need for
automated solutions. In this study, a variety of machine learning models were
applied to a CTG dataset after preprocessing to address class imbalances and
optimize feature selection. The models were evaluated based on accuracy and
ROC AUC (Receiver Operating Characteristic - Area Under Curve) scores. The
XGBoost and LightGBM models demonstrated exceptional performance,
achieving accuracies of 98.11% and 97.84%, respectively, along with near-perfect
ROC AUC scores of 0.9985 and 0.9984, indicating their ability to reliably
distinguish between the three fetal health categories: Normal, Suspect, and
Pathological. These results highlight the potential of XGBoost and LightGBM as
highly effective tools for real-time fetal health assessment, offering significant
advantages over traditional manual methods. This approach not only provides
accurate predictions but also presents a scalable and efficient solution for
resource-constrained healthcare settings. By enhancing medical decision-
making through automated fetal health monitoring, this study aims to
contribute to the reduction of preventable maternal and neonatal deaths,
particularly in low-resource environments where timely interventions are
critical.
