Recurring patterns in stationary intervals of abdominal uterine electromyograms during gestation

Abstract

Abdominal uterine electromyograms (uEMG) studies have focused on uterine contractions to describe the evolution of uterine activity and preterm birth (PTB) prediction. Stationary, non-contracting uEMG has not been studied. The aim of the study was to investigate the recurring patterns in stationary uEMG, their relationship with gestation age and PTB, and PTB predictivity. A public database of 300 (38 PTB) three-channel (S1-S3) uEMG recordings of 30 min, collected between 22 and 35 weeks' gestation, was used. Motion and labour contraction-free intervals in uEMG were identified as 5-min weak-sense stationarity intervals in 268 (34 PTB) recordings. Sample entropy (SampEn), percentage recurrence (PR), percentage determinism (PD), entropy (ER), and maximum length (L MAX) of recurrence were calculated and analysed according to the time to delivery and PTB. Random time series were generated by random shuffle (RS) of actual data. Recurrence was present in actual data (p<0.001) but not RS. In S3, PR (p<0.005), PD (p<0.01), ER (p<0.005), and L MAX (p<0.05) were higher, and SampEn lower (p<0.005) in PTB. Recurrence indices increased (all p<0.001) and SampEn decreased (p<0.01) with decreasing time to delivery, suggesting increasingly regular and recurring patterns with gestation progression. All indices predicted PTB with AUC≥0.62 (p<0.05). Recurring patterns in stationary non-contracting uEMG were associated with time to delivery but were relatively poor predictors of PTB.

Electrohysterography in the diagnosis of preterm birth: a review

“…Statistically significant differences were obtained for sample entropy between term and preterm delivery records recorded before the 26 th GA and between all term and all preterm delivery records (Fele-Zorz et al, 2008). This result was consistent with the findings of other studies, which reported that sample entropy estimated from stationary motion and labor contraction-free intervals in EHG signals using the same database was lower for preterm delivery records than term delivery records (Di Marco et al, 2014). However, when sample entropy was computed for EHG-bursts from the same database, no statistically significant difference was found between preterm and term delivery records (Horoba et al, 2016).…”

“…For patients with threatened preterm labor symptoms between 24 and 34 WG, a significant increase in approximate entropy computed in the bandwidth 0.24-4 Hz was found in patients who gave premature birth within 7 days, suggesting that the EHG signal becomes more complex as labor approaches (Lemancewicz et al, 2016). However, as the estimator of approximate entropy has been shown to be biased and highly sensitive to the number of signal samples (Ferrario et al, 2006), most authors prefer to use sample entropy, which is more independent of recording length and behaves more consistently, to characterize EHG signals (Fele-Zorz et al, 2008;Radomski et al, 2008;Garcia-Gonzalez et al, 2013;Di Marco et al, 2014;Horoba et al, 2016). Nevertheless, controversial results have been obtained for sample entropy estimated from EHG recordings.…”

“…Nevertheless, they found that the correlation dimension in the 0.08-4Hz bandwidth presented statically significant differences between term and preterm delivery records recorded before the 26 th GA, while no significant differences were found between all term and preterm delivery records (Fele-Zorz et al, 2008). Other authors analyzed the recurring patterns in stationary motion and labor contraction-free intervals in uterine EMG signals of the TPEHG DB (Di Marco et al, 2014). In comparison to term delivery records, recurrence indices such as percentage recurrence, percentage determinism and entropy and maximum length were higher in preterm delivery records, the individual predictors AUC of preterm birth being around 0.65.…”

“…In comparison to term delivery records, recurrence indices such as percentage recurrence, percentage determinism and entropy and maximum length were higher in preterm delivery records, the individual predictors AUC of preterm birth being around 0.65. The recurrence indices increased with decreasing time to delivery, suggesting regular and recurring patterns with gestation progression (Di Marco et al, 2014). Lempel-Ziv complexity has also been proposed to characterize EHG signal recorded from 60 patients with threatened preterm labor symptoms between 24 and 34 WG (Lemancewicz et al, 2016).…”

A Multivariate Multiscale Fuzzy Entropy Algorithm with Application to Uterine EMG Complexity Analysis

“…In other words, as the time of gestation progresses, the MSampEn/MFSampEn values for both term and pre-term delivery records drop, indicating higher predictability or less complexity of the signals as the delivery approaches [43]. On the other hand, the MSampEn/MFSampEn values are lower for pre-term delivery records (middle-left, middle-right and bottom-right panel of Figure 5a,d,g) regardless of the gestation duration at the time of recording, which confirms that the preterm delivery records are less complex or more predictable than the signals of term delivery records.…”