As previously shown, NIFECG can be used for two major types of analysis, namely FHR/FHRV and the morphological analysis of the extracted FECG waveform. Regarding gold-standards for the afore mentioned metrics, FHR/FHR can be compared with CTG (antepartum/intrapartum,
less accurate) and FSE (intrapartum, accurate), while for FECG morphology requires either FMCG (antepartum1) or FSE (intrapartum) as reference.
Although the Monica AN24 has not yet gained a significant foothold in the monitoring market, several studies and commercial NI-FECG equipment claim to obtain accurate FHR tracings [51, 466]. For instance, Reinhard et al.[346] has found that the NIFECG and CTG characterize FHR trace in a similar manner with regards to acceleration count, decelerations count and coincidence, variability and baseline forn= 27 subjects. Moreover, the author found an overall strong correlation (Pearson’sr= 0.91) between the techniques’ FHR traces. Similar results were obtained by a more recent study by [228], on which moderate to high correlations (r= 0.57−0.97) were foundn= 39 recordings of 20 min duration including 2 min of auditory stimulus. Another important aspect addressed by [407] was the lower rate of confusion between maternal HR and FHR achieved by NIFECG, when compared with CTG. High accuracy has also been reported in the literature when comparing FSE and NIFECG specially when using the open-source databases that counted with FSE for FQRS gold-standard.
With regards to the morphology of the signal, in Cliffordet al.[85], the authors recorded the NIFECG on n = 32 term laboring women who had FSE placed after clinical indication. They evaluated the accuracy of the FST segment extracted on the NIFECG (by an automated algorithm) against the reference FST segment extracted from the FSE (using the same automated algorithm). The root mean square error between the FST calculated by both modalities averaged over all processed segments was 3.2%, indicating that accurate extraction of the FST segment from the NIFECG may be feasible. Similarly, [344] investigated the feasibility of NIFECG FST analysis by comparing the AN24 and STAN onn= 6 pregnant women during birth. Non-invasive FST was possible in 50 % the cases due to absence of fetal T-wave. Furthermore, McDonnellet al. [274] has found a very low difference between T/R ratios obtained by STAN and the Meridian monitor forn= 27 term laboring women. Lastly, in Beharet al.[56] the authors showed the possibility to recover the FQT from the NIFECG fromn= 22 term women. The study made use of manually annotated FQT on both NIFECG and FSE, which were fused prior to comparison. The errors found between NIFECG and FSE FQT were in the range of QT annotations performed on adult ECGs [51]. All these studies are very encouraging, nevertheless a careful reader should have concerns about the reference being provided by a single cephalic lead and how the effect of the electrode positioning may affect the final results for NIFECG technique. Since the current studies are fairly limited in number, such conclusions should be considered with caution.
Considering the quality of such recordings, as early as 1995 Croweet al.[100] have qualita- tively shown the benefits and challenges in long-term NIFECG monitoring, due to the lower SNR of the fetal signal. Tayloret al.[423] reported being able to visually inspect FECG sig- nal (including P, QRS and T waves) in 80 % for this study’s population (15 pregnant women, 24-41 WOG, half of which with≥ 39 weeks) using 15 min recordings. Pieriet al.[333] made use of 400 short-term recordings (5-10 min) at different stages and pregnancy, using a FHR obtainal success rate the authors produced trends for signal quality throughout pregnancy.
1 Please note that there is a fundamental complication in using both NIFECG and FMCG simultaneously, due to the
Although little information on the subjects pathophysiological states is revealed, it is clear that around the 28thWOG NIFECG’s quality is expressively reduced. Fuchset al.[145], who used the KOMPOREL, investigated pregnant women ranging between 28 and 42 WOG and have found no correlation between the percentage of signal loss and gestational age, nor between signal loss and BMI. A more comprehensive study using the AN24 by Graatsmaet al.[163] evaluated long-term 15 h recordings of 150 preterm pregnant and 1-hour recordings of 22 laboring women, who had FSE simultaneously applied. Regarding the quality of recordings, the authors have found that 82 % the long-term recordings were of good quality (i.e. where on 60 % of the time FECG signal was present – as defined by the authors). The study also suggests that during the night period signal quality increased. According to the authors, a strong and signif- icant correlation between FSE and NIFECG’s FHR (r= 0.99) and FHRV short-term variability [106] (r= 0.79) were found during labor. Analogously, Reinhardet al.[345] assessed the signal quality of the FHR estimated fromn= 144 NIFECG and CTG recordings during the first and second stages of labor. The study showed significant better results for NIFECG during the first stage, while no difference in signal loss was found during the second stage. Associated to these results, both Graatsmaet al.[163], Van Laaret al.[439] have found no significant effect of BMI on NIFECG recording quality, however Van Leeuwenet al.[446] showed an increase in signal loss due to obesity.
Aside from the few publications herein listed and a handful of others provided by the current NIFECG commercial equipment, there is not much evidence on the ability of these devices to extract and detect fetal signals. For instance, neither Monica’s nor Meridian have published any large randomized trial to compare NIFECG recordings with other gold-standards such as CTG and FSE. Nor have they demonstrated NIFECG’s ability to improve neonatal outcome. Moreover, little is known about the quality of NIFECG recordings, particularly on earlier WOG (i.e.W OG <28). This latter analysis is important to figure out what factors aside from the vernix caseosa can influence the quality of NIFECG recordings, which would enable the technique to be used in a broader number of patients. Moreover, as previously describe, a study on FHRV is only complete if additional information about the fetal behavioral state, time of day and activity (i.e. fetal movement) is available. Similarly, to be able to analyze the FECG morphology, information about the observed projection (i.e. electrode positioning) is required. Such information could be partially obtained by obtaining a VCG representation of fetal heart, as suggested by [452].
In any manner, the fetal signal component in the NIFECG should be reliably extracted and detected before any further clinical analysis take place. Due to the more fundamental data collection and signal processing problems that are involved in NIFECG technique, the clinical analysis of the parameters that could be obtained through these recordings (i.e. FHR, FHRV and morphological analysis) are out of the scope of this thesis.