Study population

Pregnant women carrying a singleton fetus without known congenital anomalies and a gestational age between 18 and 24 weeks of gestation were included. All patients were older than 18 years and gave written informed consent prior to the fetal ECG measurement.

Patients who did not understand the Dutch language well and/or had multiple pregnancies were excluded. If CHD was found later in pregnancy or after birth, the measurement was excluded from analysis.

The following data was gained prospectively: maternal gravidity and parity, as well as obstetric and general medical history. Parents received a questionnaire three months after birth to confirm that the child was healthy and did not have any congenital diseases. We chose this three-month cut-off point as at this age, all children in the Netherlands have had their second medical check-up by a doctor, who, among other things, evaluates cardiac health using auscultation.

fECG measurements and signal processing

Singular fetal ECG measurements were performed subsequent to the fetal anomaly scan. Women lay in a semi-recumbent position to prevent aortocaval compression. To yield six channels of bipolar electrophysiological measurements, eight electrodes were placed on the maternal abdomen in a fixed configuration. Two electrodes served as common reference and ground electrodes respectively (Fig 1; [23]). Before application of the electrodes the skin was washed with water and soap after which skin preparation was performed with medical abrasive paper (Red DotTM Trace Prep, 3M Health Care, Ontario, Canada) to optimize skin impedance. Each measurement lasted around 30 minutes during which fetal orientation was ultrasonographically checked at four fixed time intervals. After training by an experienced gynecologist or sonographer the researcher determined the fetal orientation by ultrasound. The fetal orientation was determined following a step-by-step plan in which the spine was taken as an identifiable landmark. The ultrasound probe was held only in a horizontal and/or vertical position for it to be reproducible and annotations were made about the position of the probe. To evaluate the accuracy of our orientation correction, the correct ECGs/VCGs of a single participant are compared to verify their consistency.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Measurement set-up of the non-invasive fetal electrocardiogram.

Eight electrodes were placed on the maternal abdomen in a fixed configuration. Two electrodes served as common reference (Ref) and ground (Gnd). The cartesian coordinate system as used in our analyses is displayed in the bottom right corner [23].

https://doi.org/10.1371/journal.pone.0256115.g001

Fetal ECG measurements were performed with a 6-channel electrophysiological amplifier (Nemo Healthcare BV, The Netherlands) using adhesive Ag/AgCl electrodes (Red DotTM, 3M Health Care, Ontario, Canada) on the maternal abdomen. The measured electrophysiological signals were digitized at 500 Hz sampling frequency and stored on a computer for offline analysis.

This offline analysis consisted of a series of signal processing steps, aimed to suppress interferences and standardize the fetal ECG signals for fetal orientation, so that the fetal electrical heart axis could be measured. These signal processing steps have been described in more detail in Lempersz et al. 2020 [23]. In the first step of signal processing, interferences from the maternal ECG, abdominal muscles, and extracorporal sources were suppressed by an adaptive template-based method [24]. As a result, for each of the six recorded signals a fetal ECG signal was obtained, yet at relatively low signal-to-noise ratio. Because each fetus could have a different orientation with respect to the maternal abdomen and the recording electrodes placed on this abdomen, the fetal ECG signals not only changed between participants, but also within participants due to fetal movement.

The second step in the signal processing aimed to standardize for fetal orientation. To allow for such standardization, a fetal vectorcardiogram was calculated for every heartbeat first, combining the information from the six abdominal signals into a 3-dimensional fetal ECG complex [25]. This vectorcardiogram could subsequently be tracked over time, detecting fetal movements and correcting for them by rotating the fetal vectorcardiogram in 3-dimensional space. Finally, another rotation in 3-dimensional space was applied that corrected for the fetal orientation, which was assessed from intermittent ultrasound scans. For instance, if the ultrasound indicated that the fetus was in a cephalic position, the recorded fetal vectorcardiogram was rotated by 180 degrees to represent the fetal vectorcardiogram as if the fetus was in a breech position, mimicking the position used when making adult ECGs. Similarly, the fetus was rotated along the longitudinal axis as if the fetal back was orientated towards the back of the mother. The parts of the measurements of sufficient signal quality, closest to the performance of the ultrasound determining fetal orientation, were used to create the vectorcardiogram.

Finally, to enhance the signal-to-noise ratio, orientation-standardized fetal vectorcardiograms were averaged over multiple heartbeats to yield one fetal vectorcardiogram per measurement.

The orientation of the electrical heart axis was defined as the direction in which the vectorcardiogram had its maximum amplitude [25]. The latter direction was estimated as the average direction of the dominant vectors in the QRS complex, defined as the vectors from the point that the R-wave exceeded 70% of its maximum value until the point that it fell below 70% of the maximum value. The orientation of the fetal heart axis was expressed in degrees ranging from minus 180° to plus 180° and calculated in the frontal plane, where minus 90° is located superiorly.

Statistical analysis

The observed frontal angle was determined in the (x,y)-plane. The normalized coordinates () were calculated as the division of the originate coordinates (x, y) by their Euclidean norm .

We calculated descriptive statistics (median with interquartile range (IQR)) on the normalized () Cartesian coordinates. We also reported the average frontal axis with 90% prediction intervals that would function as reference values. Prediction intervals are chosen because they account for the uncertainty in estimating the population mean and the random variation of the individual values [26]. The average frontal axis is a circular mean, therefore the maximum likelihood estimate (mle) of the mean direction parameter μ of a von Mises distribution will be used as the average frontal axis.

The prediction intervals were calculated, using the lower and upper quantiles of the Von Mises distribution with the estimated parameters.

Statistical analysis was conducted with SAS (version 9.4, SAS Institute Inc., NC, USA) and R (version 3.5.3, R Foundation, Vienna, Austria). Descriptive statistics (median with interquartile ranges) were used to describe baseline characteristics, using IBM SPSS statistics version 25.0 (SPSS Inc., Chicago, Ill., USA).

Data are available upon request.

Main findings

In this paper we present reference values for the electrical heart axis calculated from our cohort of 281 healthy fetuses at mid-gestation. We found the mean electrical heart axis of the healthy fetus orientated to the right (122.68°), which is in line with the distribution of fetal cardiac muscle mass due to the unique anatomy of the fetal circulatory system and findings from previous studies [17–20, 27]. We found that the prediction intervals based on our cohort are wide, indicating a broad range wherein future observations will fall.

Strengths and limitations

The main strength of this study is the large group of participants and the low number of recordings excluded due to inadequate data quality. The latter shows that this technology has improved significantly compared to earlier reported research [28–30]. This high success rate is an indispensable characteristic for any technology to be implemented in daily practice. However, the time needed to process the recordings is at this moment the limiting factor for the NI-fECG technology, which currently still takes place offline. Therefore, results are not yet readily available during the measurement. This can be solved by automatization of the signal processing algorithms in the future which can then be incorporated in the measurement hardware. Furthermore, a few factors can contribute to inaccuracies in the correction for the fetal orientation; inaccuracies in the correction for the fetal movement and inaccurate assessment of the fetal orientation and the time at which this assessment was made. To minimize these inaccuracies, we have used the ECG data that was recorded around the time that the fetal orientation assessment was done. Possible cumulative errors in the correction for movement will therefore be limited. As an extra check, the fetal orientation was determined multiple times per recording. Within a single patient, the correct ECGs/VCGs are compared to verify their consistency, which can act as an indirect evaluation of the accuracy of our orientation correction.

Interpretation

To our knowledge, this is the first study that determines reference values for the electrical heart axis in healthy midterm fetuses. Recent advances in the signal processing algorithms have made it possible to acquire information on the fetal ECG in the antenatal period in a non-invasive manner. This makes it possible to define reference values for the electrical heart axis in healthy fetuses in mid-pregnancy.

The electrical heart axis reflects the distribution of muscle mass in the fetal heart. In the fetal circulation with its three obligatory shunts and the high resistance pulmonary and low resistance systemic circulations, the right ventricle is dominant and pumps about 60% of the cardiac output. As a consequence the muscle mass of the right ventricle is greater than that of the left ventricle and this results in greater amplitude of depolarization together with decreased speed of depolarization on the right side [31]. Our results confirm this right oriented electrical heart axis in healthy fetuses. The next step towards determining the use of this parameter for screening purposes is to define the electrical heart axis in fetuses with CHD.

Fetal electrocardiography is an easy to use, non-invasive, safe technology with a minimal burden for the pregnant women. Further research towards the electrical heart axis in fetuses with different types of CHD is necessary to determine which defects are associated with a deviated fetal electrical heart axis. Then the NI-fECG could be performed in addition to the fetal anomaly scan around the 20^th week of gestation as part of prenatal screening after automatization of the signal processing of the recording. A point of attention is the broad distribution of the electrical heart axis found in our cohort of healthy fetuses in mid-pregnancy. This resulted in wide predictions intervals [-25.6°; 270.9°] making the use of the electrical heart axis alone as a parameter for the screening of CHD less suitable. Future research towards ECG waveform and ECG intervals may add to the development of additional ECG parameters which could further enhance the prenatal detection of CHD.

Conclusion

Our results confirm that the mean electrical heart axis of healthy fetuses around mid-gestation is oriented to the right. The wide prediction interval for the frontal heart axis found in our cohort, is unfavorable for future implementation of this method for screening purposes. Further research towards the electrical heart axis in fetuses with CHD as well as additional ECG waveform and intervals may elucidate the role of fetal ECG as a screening parameter for the detection of CHD.