•  
  •  
 

Abstract

Fetal health assessment is crucial for ensuring the well-being of both mother and fetus during pregnancy. Accurate monitoring can reduce child and maternal mortality rates, particularly in low- and middle-income countries where these rates are higher. This research aims to enhance fetal health classification using advanced machine learning techniques and Explainable Artificial Intelligence (XAI) to address the gaps in predictive accuracy and model transparency. A dataset of 2,126 records with 22 features from Cardiotocography (CTG) readings. The dataset was balanced using Synthetic Minority Over-sampling Technique (SMOTE) to handle class imbalances. Various machine learning algorithms, including LightGBM, XGBoost, and Random Forest, and employed Pearson's Correlation for feature selection were implemented. Shapley values were used to ensure model interpretability. LightGBM achieved the highest accuracy at 95.9%, followed by XGBoost at 95.5%. Feature importance and SHAP analysis revealed features that are critical for accurate predictions. Our study also demonstrates that combining machine learning with XAI can drastically improve fetal health monitoring by providing interpretable models. Ultimately, this contributes to more informed decision-making in fetal health monitoring and supports global efforts to reduce maternal and neonatal mortality in line with the Sustainable Development Goals.

Keywords

Cardiotocography, Ensemble Model, Fetal Health, SMOTE, XAI

Subject Area

Computer Science

Article Type

Article

First Page

2469

Last Page

2478

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS