Detecting labor using graph theory on connectivity matrices of uterine EMG

Area of Science:

• Biomedical Engineering
• Signal Processing
• Computational Mathematics

Background:

• Premature labor is a critical health issue globally.
• Current methods struggle to reliably distinguish true labor from Braxton Hicks contractions.
• Accurate discrimination is essential for timely intervention and improved maternal-fetal outcomes.

Purpose of the Study:

• To explore the utility of graph theory applied to uterine EMG signals.
• To enhance the differentiation between normal pregnancy contractions and preterm labor.
• To investigate novel signal processing techniques for improved diagnostic accuracy.

Main Methods:

• Application of graph theory techniques to multi-electrode uterine EMG signals.
• Analysis of synthetic and real uterine EMG data.
• Utilized a low-pass windowing preprocessing step for signal enhancement.
• Compared graph parameters between pregnancy and labor states.

Main Results:

• Graph theory parameters showed differences between synthetic pregnancy-like and labor-like graphs.
• The same parameters effectively differentiated real uterine EMG signals from pregnancy and labor.
• A low-pass windowing preprocessing step significantly improved discrimination accuracy.
• Uterine EMG graphs transitioned from random characteristics during pregnancy to a more organized, small-world network structure during labor.

Conclusions:

• Graph theory analysis of uterine EMG signals offers a promising approach for distinguishing true labor from pregnancy contractions.
• The findings suggest a potential for improved diagnostic tools for preterm labor.
• Signal preprocessing, particularly low-pass windowing, is crucial for optimizing this method.