Preterm delivery, defined as delivery before 37 weeks of gestation, constitutes a major problem in terms of neonatal mortality, morbidity, and healthcare costs [
The sequence of contraction and relaxation of the uterus results from a cyclic depolarization and repolarization of its smooth muscle cells in the form of action potentials (APs). APs occur in bursts; they arise in cells that act as pacemakers and propagate from cell to cell through gap junctions [
The previous literature demonstrated that the EHG has great potential for monitoring labor, predicting labor time, and discriminating between physiological uterine activity and contractions leading to (preterm) delivery. Therefore, analysis of the EHG can support timely treatment of preterm labor [
Prior to delivery, the increased connectivity among cells also increases propagation, which can be assessed by estimating the conduction velocity (CV) from the EHG [
Previous research mainly focused on the methods for measuring the CV [
However, automated CV analysis entails a number of scientific challenges, namely, automatic detection of contractions, estimation of amplitude and direction of the CV vector, and exclusion of signals that are not related to propagating APs.
This study investigates the feasibility of a new automated approach for the analysis of the EHG CV for detecting imminent delivery. Our approach integrates previously validated EHG-based methods for contraction detection and automated analysis of the CV in two dimensions using a high-density electrode grid.
A prospective observational cohort study was performed at the Maxima Medical Center Veldhoven, The Netherlands. Approval from the local medical ethical board was obtained and all the included women provided written informed consent for study participation. Patients with singleton pregnancies were enrolled, presenting with at least 3 contractions in 30 minutes, which were either perceived by the patient or visible on the external tocogram. Both term patients (gestational age
Measurements from all enrolled patients were obtained using a measurement setup as shown in Figure
Measurement setup showing the position of all abdominal sensors.
The recording of these signals was performed using a Refa multichannel amplifier (TMS International, Enschede, The Netherlands), with a patient ground on the hip. Simultaneously, a tocodynamometer was used as reference for contraction detection. For the same reason, the time instants at which the patient felt a contraction were annotated.
Here, a synthetic overview of the methodology used for the analysis is given. For further details, we refer to [
The CV vector was identified during the contraction periods. Differently from previous studies, where contractions were annotated manually, we automatically derived an initial estimation of onset and duration of contractions. To this end, an estimate of the internal uterine pressure (IUP) was derived from the bipolar EHG signal. Based on a validated method [
Of the contractions selected by EHG signal analysis, only those that were visible on the external tocogram or concurred with annotations of contractions as felt by the patient were eventually selected for further analysis.
In the signal segments selected as contractions, the CV vector was identified in two dimensions from the
The method for analyzing the CV. The EHG is recorded using a
The use of different weighting strategies of the derived cost function was introduced in [
Segments with a calculated CV value above 30 cm/s, which are significantly higher than the physiological values reported in the literature [
Patients delivering within 24 hours after the measurement were classified as labor group and those delivering outside this time window were classified as nonlabor group. CV and propagation path were compared between these groups. In order to be independent of the number of analyzed segments and contractions per patient, an average CV vector was identified for each analyzed contraction and subsequently the average CV vector for each patient was determined. The Shapiro-Wilk test was used to test for a normal distribution of the estimated values of CV vector amplitude. Levene’s test was applied to test for equal variances in the labor and nonlabor groups. Finally, an independent samples
Twenty-two patients were included in the study, of which 7 were preterm. Nine patients delivered within 24 hours and were classified as labor group. Table
Patient characteristics.
Labor | Nonlabor | |
---|---|---|
Number of patients | 9 | 13 |
Gestational age (weeks + days)1 | 31 + 1–40 + 4 (37 + 2) | 26 + 2–41 + 3 (36 + 1) |
Preterm | 2 | 5 |
Nulliparous | 4 | 8 |
Age1 | 17–36 (27.9) | 16–36 (27.8) |
BMI1 | 22–42 (28.2) | 24–34 (26.8) |
Hours to delivery1 | 1–10 (6) | 27–1488 (255) |
Next to the HD electrode grid, the EHG signals of eight electrodes (one column) are shown during a contraction. The five images at the bottom show an interpolated 2D representation of a single EHG pulse propagating from top to bottom.
In total, 64 contractions were analyzed. Figure
On the left, a boxplot shows the average amplitude of the CV vector for the labor and nonlabor groups. The median value is displayed with a horizontal line. On the right, a scatter plot shows the individual average amplitudes of the CV vector for both groups. The horizontal lines represent the average value for the labor and nonlabor groups.
This study investigates the feasibility of a new automated approach for the analysis of the EHG CV for detecting imminent delivery. Our approach integrates validated EHG-based methods for contraction detection and automated analysis of the CV in two dimensions using a high-density electrode grid [
The measurements were performed in a diverse group of patients featuring both term and preterm patients admitted for varying reasons. The common denominator was that all patients had palpable and measurable contractions. The emphasis was placed on investigating the feasibility of automated CV analysis, in order to open the way to future clinical studies and applications based on this parameter as diagnostic tool for imminent (preterm) birth. The assumption here is that comparable changes in conduction properties can be observed in contractions leading to preterm and term delivery. In follow-up studies, it will be important to have a consistent group of patients presenting with premature contractions and who are considered for treatment with tocolytics based on gestational age and clinical parameters.
In this study, additional data other than the EHG signal was used for detecting contractions, namely, an external tocodynamometer and annotations of subjectively perceptible contractions by the patient. This was chosen to achieve a more robust distinction between uterine activity and measurement artifacts. In future work, a fully automated selection of contraction segments and analysis of CV should be pursued. However, while an automated method ensures reproducibility and should therefore be preferred for everyday clinical use, visual inspection might be required in a preliminary phase for discriminating uterine activity from noise and excluding artifacts and signals that do not propagate linearly, from the analysis.
Noteworthily, identification of the EHG CV vector using the present methods implies the assumption that the signal does propagate and that propagation is linear. While it is reasonable to hypothesize that the linearity of the propagation could be a discriminative parameter for predicting imminent delivery in itself, several aspects related to the evolution from pregnancy to labor are not yet fully understood and need further dedicated research [
Finally, another novelty of this study is the use of a 64-channel high-density electrode grid for recording the EHG. Due to the a priori unknown AP direction of propagation, the bidimensional arrangement of the electrodes on the grid permits estimating all the possible CV directions along the abdominal plane parallel to the abdominal surface. Furthermore, due to the grid dimensions, planar wave propagation could be assumed and the small interelectrode distance enables following the same spike (action potential) from one electrode to the other [
In agreement with previous studies, our results show that the CV vector amplitude is a promising parameter for predicting imminent (preterm) delivery. Automated estimation of this parameter from the EHG signal is feasible and should be regarded as an important prerequisite for future clinical studies and applications in this context. Therefore, these results open the way to future studies on the accuracy of EHG parameters, such as the CV, for timely and accurate diagnosis of imminent preterm delivery.
The authors declare that none of them have a conflict of interests.
The authors declare that the source of financial support was from ERASysBio European Research Project/Biophysical Modeling of the Uterine Electromyogram for Understanding and Preventing Preterm Labor: