Uterine EMG analysis: a dynamic approach for change detection and classification

Area of Science:

• Biomedical Engineering
• Signal Processing
• Obstetrics and Gynecology

Background:

• Preterm birth detection remains a clinical challenge.
• Uterine electromyogram (EMG) signals contain crucial information about labor.
• Characterizing uterine EMG events is key to predicting preterm birth.

Purpose of the Study:

• To develop and validate an automated method for detecting and classifying uterine EMG events.
• To improve the accuracy of preterm birth prediction through signal analysis.
• To establish a robust algorithm for real-time uterine EMG monitoring.

Main Methods:

• Utilized a dynamic change detection approach assuming piecewise stationarity in uterine EMG.
• Employed dynamic cumulative sum (DCS) with wavelet transform for multiscale signal decomposition.
• Implemented unsupervised classification using variance-covariance matrices and neural networks for event labeling.

Main Results:

• The proposed DCS and multiscale decomposition method efficiently detects frequency and energy changes in uterine EMG.
• The detection-classification-labeling algorithm achieved over 80% accuracy in event identification and categorization.
• Satisfactory performance was observed across various stages of gestation.

Conclusions:

• The developed algorithm offers a promising tool for the detection and classification of uterine EMG events.
• This method has the potential to significantly enhance preterm birth detection and management.
• The approach demonstrates robustness and high accuracy in characterizing complex uterine EMG signals.