Electrohysterography in modern obstetrics: Advances in signal processing, machine learning, and clinical applications

Electrohysterography (EHG) represents a promising computational approach for non-invasive monitoring of uterine activity during pregnancy and labor. This review summarizes the advancements in signal processing techniques and machine learning algorithms that have been applied to enhance the utility of EHG. Key topics include the extraction and analysis of uterine electrical signals, classification of contractions, and prediction of obstetric outcomes such as preterm and labor/non-labor states. The review emphasizes computational methodologies for signal processing and extraction, including empirical mode decomposition or wavelet transform, and for data classification, such as neural networks or support vector machine, highlighting their performance and limitations. Despite significant progress, challenges persist, such as the lack of standardized protocols, limited datasets, and inconsistent evaluation and annotation metrics, which hinder broader clinical adoption. The integration of additional clinical markers, simultaneous monitoring of maternal and fetal health, and the development of wearable systems for telemedicine present exciting opportunities for future research.