An unsupervised classification method of uterine electromyography signals using wavelet decomposition

Area of Science:

• Biomedical Engineering
• Signal Processing
• Obstetrics

Background:

• Uterine contractions are crucial indicators of labor and pregnancy progression.
• Electromyography (EMG) signals reflect the electrical activity of uterine muscles during contractions.
• Variability in contraction frequency necessitates advanced analysis techniques.

Purpose of the Study:

• To classify uterine contractions based on their electromyography (EMG) signal characteristics.
• To develop an objective method for analyzing and categorizing uterine contraction patterns.
• To investigate the relationship between EMG signal frequency content and contraction types.

Main Methods:

• Utilized wavelet decomposition to extract key parameters from uterine EMG signals.
• Employed an unsupervised statistical classification method, specifically the Fisher test, for event categorization.
• Applied Principal Component Analysis (PCA) projection to visualize and confirm the classified groups.

Main Results:

• Successfully classified uterine contractions into distinct groups based on their frequency content.
• Demonstrated that frequency characteristics of EMG signals vary significantly between different contraction types.
• PCA effectively highlighted the separation of contraction groups derived from the Fisher test classification.

Conclusions:

• Uterine contractions can be reliably classified using EMG signal analysis.
• Wavelet decomposition and statistical classification offer a robust approach to understanding contraction dynamics.
• This method provides a foundation for improved monitoring and understanding of uterine activity during pregnancy.