Dependency of magnetocardiographically determined fetal cardiac time intervals on gestational age, gender and postnatal biometrics in healthy pregnancies

FMCGs were acquired using a 61 channel biomagnetometer (Magnes 1300C, 4D Neuroimaging, San Diego) in 162 cases. The sensing channels are arranged as 4 concentric rings around a central channel with an overall diameter of 32.4 cm and an area of coverage of ca. 800 cm². A configuration of 11 reference coils is used to detect ambient noise and intrinsic system noise is given as being less than 10 fT/Hz^1/2 (fT: femtotesla) for frequencies > 5 Hz. A further 68 FMCGs were recorded using a 37 channel system (Krenikon, Siemens, Erlangen). This system consists of sensing coils configured as 1st order gradiometers (baseline 7 cm) in a hexagonal grid with an overall diameter of 19 cm (area of coverage: ca. 283 cm²). System noise is given as less than 30 fT/Hz^1/2 for frequencies > 2 Hz and 10 fT/Hz^1/2 for frequencies > 10 Hz. Previous work has shown that the estimation of CTI is comparable for these two systems [10]. For data acquisition the mothers were in a supine position with the biomagnetometer placed as close to the abdomen as possible while avoiding skin contact. The 61 channel system was positioned symmetrically over the abdomen, the lower rim over the pubis and the upper rim approximately over the xyphoid. The smaller 37 channel system was positioned such that channels at the center of the sensor displayed high amplitude fetal QRS complexes. Data were usually recorded for 5 minutes at a sampling rate of 1 kHz and with a bandpass of 1–200 Hz. To reduce the effects of external noise, all measurements were performed in a standard shielded room (AK3b, Vacuumschmelze, Hanau). Recordings were generally performed between 10 AM and 4 PM.

In each set of data, maternal heart beats were identified by correlation to a maternal QRS signal template and the maternal PQRST signal components were then digitally averaged and subsequently subtracted from the signal traces. In the resulting traces, fetal beats were similarly identified on the basis of a representative fetal QRS signal template. Fetal PQRST courses to be averaged were chosen on the basis of a good correlation (r ≥ 0.90) to the template (generally»300 beats). In the averaged data, the onsets (or endpoints) of the P wave, QRS complex and T wave were defined as the visually identifiable first (or last) deviation from the signal's baseline in any one channel out of all channels available for evaluation [11]. Due to the low amplitude of the T wave and possible DC drift, its onsets and end were determined on the basis of identifiable changes in amplitude at appropriate latencies. All channels were displayed individually avoiding overlap and within a fixed time window (20 ms/cm) and a defined range of amplitudes (0.1–0.5 pT/cm for the P and T wave, 0.5–1.0 pT/cm for the QRS complex). The onsets and ends were independently determined by two experienced investigators and their results were compared. The final value for each time point was set accordingly: in cases of differences > 2 ms both investigators re-evaluated and corrected the results, otherwise the results were averaged. These time points were then used to calculate the duration of the CTI. We determined the consecutive, non-overlapping (disjunct) intervals as follows: P wave = P_end - P_onset, PQ interval = QRS_onset - P_end, QRS complex = QRS_end - QRS_onset, ST segment = T_onset - QRS_end and T wave = T_end - T_onset. Furthermore, the following composite CTI were determined: PR interval = P wave + PQ interval and QT interval = QRS complex + ST segment + T wave (see Figure 2). Also, the rate corrected QT interval (QTc) was calculated according to Bazett's formula [18]. In some traces, signal quality did not permit the unequivocal identification of the timing of an event and the corresponding CTI were not determined.

The dependency on gestational age was assessed for the CTI over all data sets using regression analysis. The same dependency in the male and female subgroups was similarly examined and compared. In order to examine the relationship between postnatal biometric data (birth weight, head circumference and crown-heel length) and the CTI, we examined the dependency of those intervals which clearly correlated to development. For this, only data sets in which the CTI were determined within the last 7 prepartum days were considered.

Values describing the subject and FMCG acquisition groups were expressed as means ± standard deviation. Simple linear as well as nonlinear regression was used to estimate the parameters governing the change of the CTI over gestational age. For all regressions, the dependencies due to repeated measurements were considered by performing a robust regression using a modified Huber/White/Sandwich estimator of variance [19]. For any interval, the model that demonstrated the best fit, based on the coefficient of determination, was chosen as most appropriate. For those intervals displaying a clear dependency on gestational age with respect to the best fitting model, estimated values and 90% prediction intervals were calculated on a weekly basis. For those intervals lacking a clear dependency, measures of location were calculated overall. To examine the differences between male and female CTI at a specific week of gestation, all values of each gender were projected to that week on the basis of each sex's best fitting model and the projected group values compared using Mann-Whitney U test. The dependency of the CTI on biometric parameters was examined using simple linear regression. Statistical significance was assumed at a level of 0.05.

The CTI showed varying dependency on gestational age: Figure 3 shows that, of the contiguous, disjunct intervals, in particular the duration of the P wave and QRS complex increased clearly over time whereas the PQ interval showed a slight tendency to decrease with age. The ST segment and T wave duration demonstrated a wide spread without a clear trend. The composite intervals were characterized by a modest positive relationship to week of gestation, primarily reflecting the influence of the depolarization times. Regression analysis confirmed the visual impression (Table 1) with statistically significant slopes for all intervals except the ST segment and the T wave. The model fits were especially good for the P wave and the QRS complex. Implementing non-linear models improved the fit for the PQ and PR interval as well as the QRS complex.

For the intervals showing the clearest changes over time (P wave and QRS complex) we calculated various measures of location on the basis of the best fitting models for each gestational week from the 17^th to the 42^nd (see Table 2, Table 3). Although the dependency on age was not as pronounced, we did the same for the PR and PQ intervals (Table 2). For the remaining intervals, the scatter of the data was too large to permit a meaningful differentiation according to week of gestation (see Fig. 3) and the numerical description of the data is given without reference to gestational age (Table 4).

With respect to fetal gender, most intervals showed no obvious differences in duration (Figure 3). However, the duration of the QRS complex seemed to be shorter in the females, particularly after the 30^th week of gestation. As this suggested a different development between the two sexes, we examined the dependency on age separately for these subgroups using different models. We found that the best fitting model for the males was logarithmic (ln(QRS) = 0.594*ln(week) + 1.852, r ² = 0.64, p < 0.001) and hyperbolic for females (QRS = -646*(1/week)+68, r ² = 0.71, p < 0.001). Using these respective models to adjust for different weeks of gestation showed that QRS duration differed significantly between gender from the 31^st week onward. We thus used these model to calculate values for the QRS complex separately for male and female subjects (Table 3).

As the P wave and QRS complex showed a distinct relationship to gestational age reflecting fetal growth, we assumed that an association to postpartum biometric parameters might exist. We therefore examined the duration of these two intervals with respect to birth weight, head circumference and crown-heel length in those fetuses in whom MCG recording were obtained within 7 days of parturition. None of the intervals showed a statistically significant relationship to the biometric data (Table 5). However a trend was apparent for head circumference, especially with respect to the QRS complex.