An Automatic Method for the Segmentation and Classification of Imminent Labor Contraction From Electrohysterograms

Area of Science:

• Obstetrics and Gynecology
• Biomedical Engineering
• Perinatal Medicine

Background:

• Preterm birth is a leading cause of perinatal morbidity and mortality, with current diagnostic methods lacking specificity.
• Despite clinical advancements, the preterm birth rate remains high, and early diagnosis is challenging.

Purpose of the Study:

• To develop and validate an electrohysterogram (EHG) processing technique for classifying pregnant women at risk of preterm delivery.
• To enhance the sensitivity and specificity of early preterm labor diagnosis using non-invasive EHG recordings.

Main Methods:

• Utilized non-linear correlation analysis for extracting uterine contraction bursts from EHG signals.
• Employed Gaussian mixture models for classifying patients based on computed EHG features.
• Tested the method on a dataset of 68 EHG recordings from women hospitalized for preterm delivery threat.

Main Results:

• Achieved automatic segmentation of contractions with high performance.
• Reported classification sensitivity, specificity, and accuracy of 80.7%, 76.3%, and 76.2%, respectively.
• Demonstrated promising results comparable to gold standards but with a non-invasive approach.

Conclusions:

• EHG recordings can reliably diagnose imminent labor, offering a non-invasive alternative to current methods.
• This technology has the potential to reduce over-medication and improve care for high-risk pregnancies.
• The non-invasive and home-monitoring potential of EHG can significantly benefit perinatal care.