The present work seeks to determine if a particular non-linear analytic method is effective at quantifying uterine electromyography (EMG) data for estimating the onset of labor. Twenty-seven patients were included, and their uterine EMG was recorded non-invasively for 30 min. The patients were grouped into two sets: G1: labor, N=14; G2: antepartum, N=13. G1 patients all delivered spontaneously within 24 h of recording while G2 patients did not. The uterine electrical signals were analyzed offline by first isolating the uterine-specific frequency range and then randomly selecting "bursts" of uterine electrical activity (each associated with a uterine contraction) from every recording. Wavelet transform was subsequently applied to each of the bursts' traces, and then the fractal dimension (FD) of the resulting transformed EMG burst-trace was calculated (Benoit 1.3, Trusoft). Average burst FD was found for each patient. FD means for G1 and G2 were calculated and compared using t test. FD was significantly higher (P<0.05) for G1: 1.27±0.03 versus G2: 1.25±0.02. The wavelet-decomposition-generated fractal dimension can be used to successfully discern between patients who will deliver spontaneously within 24 h and those who will not, and can be useful for the objective classification of antepartum versus labor patients.
ASJC Scopus subject areas
- Biomedical Engineering
- Computer Science Applications