Machine Learning- Based Fetal Risk Classification Using Intrapartum Cardiotocography

Cardiotocography (CTG) is widely used during labour to monitor fetal well-being, but its interpretation remains challenging due to the complexity of fetal heart rate patterns and variability in clinical assessment. This project presents a machine learning framework for automated fetal risk classification using the CTU-UHB intrapartum CTG dataset. Patient-level labels were generated from neonatal outcome measures, and 55 clinical and temporal features describing fetal heart rate and uterine contractions were extracted to create an interpretable representation of each recording. Five machine learning models were evaluated using repeated cross-validation and predefined train-test splits. The results demonstrate that incorporating temporal information improves predictive performance while maintaining model interpretability. The proposed framework provides a transparent and reproducible foundation for automated fetal risk classification and future clinical decision-support systems.