Predictive Obstetrics: Electrohysterogram-Based Detection of Preterm Labor.

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare
• Obstetrics and Gynecology

Background:

• Preterm birth is a leading cause of infant mortality and morbidity.
• Accurate prediction of preterm labor is crucial for improving maternal and neonatal outcomes.
• Electrohysterography (EHG) offers a non-invasive method for monitoring uterine activity.

Purpose of the Study:

• To classify term and preterm labor records using machine learning algorithms.
• To evaluate the performance of decision trees, subspace k-nearest neighbors, and neural networks for preterm birth prediction.
• To assess the clinical relevance of EHG and machine learning in obstetric practice.

Main Methods:

• Utilized three machine learning algorithms: decision trees, subspace k-nearest neighbors, and a trilayered neural network.
• Classified electrohysterography (EHG) data to distinguish between term and preterm labor.
• Evaluated algorithm performance using accuracy, sensitivity, positive predictive value, and F1-score.

Main Results:

• Decision trees demonstrated the highest classification accuracy (85.56%) on the imbalanced dataset.
• All tested algorithms faced challenges in classifying the minority class (preterm births).
• Performance decreased on synthetically balanced data, suggesting issues with data quality.

Conclusions:

• Machine learning applied to EHG data can aid in predicting preterm labor.
• Decision tree algorithms show potential for improving preterm birth risk assessment.
• Further research is needed to address challenges with imbalanced data and enhance prediction accuracy for clinical application.