Comparing achievability and reproducibility of pulsed wave Doppler and tissue Doppler myocardial performance index and spatiotemporal image correlation annular plane systolic excursion in the cardiac function assessment of normal pregnancies

Introduction

Evaluation of fetal cardiac function involves quantifying chamber shape and size, cardiac volume and stroke volume, contractility and relaxation [1], to detect adaptation to strain [2], [3], [4]. The fetal and adult cardiovascular systems differ [5] with the fetus having parallel circulations, shunts, right heart dominance and differing sensitivity to loading conditions from the adult [6], [7], [8]. The fetal heart has a longitudinal myocardial fibre arrangement on the right and circumferential on the left [9], 10]. Several Doppler tools have been proposed for functional fetal cardiology.

The Myocardial Performance Index (MPI) is a pulse-wave Doppler-derived index of global myocardial function, relying on time intervals, therefore having the advantage of being relatively independent of precise imaging, anatomy, ventricular geometry and heart rate [11]. The Modified-MPI variation (Mod-MPI) introduced valve clicks to delineate time intervals better [12] although calliper placement varies across the literature [12], [13], [14], [15]. We have previously shown that automation for both left and right MPI removes the dependence on observer calliper placement, expertise and subjectivity [16], 17].

Tissue Doppler imaging (TDI) has been proposed as a robust tool for the detection of subclinical cardiac dysfunction with improved reproducibility over conventional Doppler [18], and unlike PW-TDI, the right TDI-MPI can be acquired in a single acquisition [19]. It is also possible to measure the tricuspid/mitral annular plane systolic excursion (TAPSE/MAPSE) or maximal longitudinal displacement of the valve annuli [20], 21]. TAPSE/MAPSE may also be measured using spatiotemporal image correlation (STIC) to overcome challenges arising from the orientation of the fetal cardiac longitudinal axis, through post-processing multi-planar reconstruction and manipulation [22].

Whilst these three independent modalities (PWD-MPI, TDI-MPI, and TAPSE/MAPSE) have been applied to cardiac function evaluation, we aimed to perform a multi-modality global cardiac function evaluation comparing their achievability and reproducibility.

Methods

This prospective, cross-sectional study recruited patients from a tertiary hospital antenatal clinic with local ethics committee approval (SESLHD HREC Reference 13/320).

Seventy-nine patients were consecutively recruited between February and November 2015, and within each case, the best images and volumes were selected. Women eligible to be included in the study were older than 18 years with an uncomplicated singleton pregnancy between 22 and 39 gestational weeks and a normal morphology scan. Women with known complications, whether maternal (such as hypertension, pre-eclampsia, diabetes) or fetal (such as IUGR or structural anomalies), were excluded. Women with insufficient English to consent were also excluded.

The necessary sample size for comparison of intra-observer repeatability comparing n=3 replicates per subject (for all three modalities) was calculated to be k=39.6, aiming for intra-observer ICC reliability of 0.7 with a minimum ICC>0.5 acceptable [23] and with a type I error of 0.05 and a type II error of 0.2 [24]. A larger sample size was planned to ensure sufficient sets of optimal scans.

One experienced operator performed the ultrasound scans. Following routine biometry and fetal well-being measurements, STIC volumes for TAPSE and MAPSE and PWD and TDI plots for MPI measurements were acquired, as depicted in Figure 1. Left PWD-MPI was acquired in the four-chamber view in an apical or basal position [12], 25]. Right PWD-MPI was acquired in two planes (tricuspid inflow and pulmonary outflow plots) [26]. Right and left TDI-MPI were acquired in the four-chamber view at either an apical or basal position [27], 28], and then multiple STIC volumes were acquired and saved to ensure at least three were of adequate quality for analysis. The insonation angle was maintained as close to the apex directly up or down, with a zoom into the region of interest at a minimal volume angle of 15° and an acquisition time of 7.5 s. Volumes with fetal movement were excluded and repeated, and total scanning time was recorded for each patient.

Figure 1:

Study design.

All MPI waveforms and STIC volumes were reviewed offline, and the three clearest cardiac cycles in every waveform (TDI-left, TDI-right, PWD-left, two PWD-right plots) and the best three STIC volumes were chosen for each case by the lead researcher. TDI images were classified for quality (scale 1–3) [27]. The angle of insonation, defined as the angle between the ultrasound beam axis (from the sector tip to the distal field) and the cardiac axis (along the ventricular septum in the 2D image or STIC volume), was recorded. Individual 2D scans were excluded if: i) the time intervals (TDI) or valve clicks (PWD) were indistinguishable; ii) the automation program failed to identify clicks (PWD); iii) the measured difference in heart rate between the two PWD-right Doppler plots was greater than 10 beats per minute.

Both left and right TDI-MPI and Left PWD-MPI values were calculated as the sum of the isovolumetric time intervals divided by the ejection time ( ICT + IRT ET ) . Right PWD-MPI was calculated as the ‘a’ interval (measured from the tricuspid flow plot) minus the ‘b’ interval (measured from the pulmonary flow plot) divided by the ‘b’ interval a − b b [28], 29]. Left and right PWD-MPIs were measured automatically using locally developed and verified algorithms [16], 17]. Inconsistency of waveform patterns with TDI-MPI and difficulties with describing clear landmarks for delineating time intervals precluded similar automation, so a single experienced operator manually measured the time intervals [30], [31], [32], [33], [34], [35]. For the STIC volumes, the cine-like heart cycle loop speed was adjusted to a lower speed (25–50 %) to allow a better appreciation of the image details.

Data was anonymised, and two experienced operators evaluated all the volumes. Intra-observer repeatability was analysed by comparing the same operator’s measurements of three different volumes from the same patient. Inter-observer repeatability was analyzed by comparing both operators’ measurements of volume no. 1 from each patient. Data was collected using Excel 2010 (Microsoft Office, Redmond, Washington, USA) and statistical analysis using SPSS version 23.0 (IBM, Armonk, New York, USA). Descriptive statistics (mean, minimum and maximum) were determined for relevant characteristic variables. Achievability was measured as a percentage success rate of achieving the required measurements. TAPSE and MAPSE values were compared using a paired T-test. Intra-and inter-observer variability were assessed with intraclass correlation coefficients (ICC) and 95 % Confidence Interval (CI), using a two-way random model using single measurements for absolute agreement [36]. Finally, a sub-group analysis was performed by dividing the study group to three sub-groups according to their gestational age: 22–27 weeks, 28–33 weeks and 34–39 weeks.

Results

Seventy-nine women with uncomplicated pregnancies were recruited and scanned. Key demographic variables are summarised in Table 1. All growth and well-being ultrasound parameters, including estimated fetal weight, Doppler studies, and liquor, were reported to be within the normal range. The total scan time was recorded with a median of 38 min and an interquartile range of 19 min (Q1=32 and Q3=51 min).

Successful acquisition rates were 86 and 92 % for left and right TDI-MPI, respectively, 87 % for both left and right PWD-MPI, and 94 % for both TAPSE and MAPSE. To minimise bias, all 222 volumes appeared randomly in the study data set for the two operators to analyse.

Descriptive statistics and 95 % CIs, respectively, for the left and right TDI-MPI and PWD-MPI values and their time intervals are summarised in Table 2. Descriptive statistics for TAPSE and MAPSE are summarised in Table 3, with mean TAPSE values being significantly higher than MAPSE (difference 1.54 mm, p<0.0001; 95 % CI 1.32–1.75, SD 0.11).

TAPSE and MAPSE values increased throughout gestation, as shown in Figure 2. In contrast to our group’s previous publication [37], there was no relationship between gestational age or fetal heart rate and either PWD or TDI MPI values. There was a poor correlation between PWD-MPI and TDI-MPI, as we have shown previously [38] Table 4 shows the repeatability for the left and right sides of the fetal heart for PWD-MPI, TDI-MPI, MAPSE, and TAPSE, and Table 5 shows the three sub-groups analysed according to gestational age.

Figure 2:

Tricuspid annular plane systolic excursion and mitral annular plane systolic excursion absolute values by gestational age.

Discussion

This study provides the first quantitative comparison between three novel fetal cardiac function modalities for their achievability and repeatability. This could potentially lead to the integration and optimisation of techniques for fetal cardiac evaluation.

Achievability rates were high for all three modalities for both sides of the heart and similar when sub-grouped by gestational age (all above 86 %). The PWD-MPI automation algorithm used in this study successfully calculates MPI (87 % of the cases) when it clearly locates the four valve clicks, effectively excluding poor-quality waveforms. Manual calliper placement in the same circumstance might have estimated the appropriate location, introducing more flexibility and variability. Lower-quality TDI scans, which were manually measured but were nevertheless deemed visually poor, were included in the analysis. Exclusion of suboptimal TDI scans by applying similar criteria to TDI-MPI would cause a significant reduction of the acquisition rate for both left and right TDI MPI (from 86 to 47 % and from 92 to 70 %, respectively). This indicates that using TDI to measure MPI in fetuses is easily achievable, though the waveform quality has considerable variability.

TAPSE and MAPSE high achievability rates (94 %) may have been a result of the novel STIC acquisition settings chosen for this study, designed to optimise annuli visualisation and favouring high-quality images. An optimal 2D image of the four-chamber view of the heart at any orientation was obtained, yet at an insonation angle as close to an apex up or apex down position. A region of interest encompassing the fetal heart was selected [39] and STIC acquisition parameters were set to the minimum volume angle of 15° and a minimal acquisition time of 7.5 s. This narrow angle maximizes the frame rate during the acquisition, improving the temporal resolution of the volume dataset, and enhancing image quality. The selected short acquisition time, minimizes the likelihood of artefacts related to fetal movement, while having the minor expense of spatial resolution [40].

As we have previously shown, PWD-MPI and TDI-MPI gave statistically significantly different and inconsistent values for both sides of the heart, with TDI-MPI measuring greater values for the left ventricle and lower values for the right, potentially reflecting the difference between measuring blood flow and myometrial contraction [27]. Variation in the TDI waveform and lack of definitive landmarks also contributes to time interval measurement inconsistency [19], 31], 41].

Fetal right heart dominance may be a reason for higher TAPSE values than MAPSE. An alternative explanation could lie in the twisting movement of the left muscle fibres compared to the more vertical motion on the right side generating a shorter vertical excursion of the mitral annulus [42].

Whilst our automated algorithm for PWD-MPI has previously shown a perfect ICC of 1.0 for the repeated measurement of the same cardiac cycle [16], 17], in this study PWD-MPI was measured across three cardiac cycles representing variability in repeat measurements completed on a single fetus during similar circumstances. This was lowest (ICC 0.69) for the right side of the heart, likely due to the introduction of variability while using two cardiac planes for recording tricuspid inflow and pulmonary outflow, adding to the underlying physiological Doppler variability.

For the right side of the fetal heart, TAPSE showed the best repeatability with inter-observer ICC of 0.86 and average intra-observer ICC of 0.73, potentially a consequence of the more distinct ultrasound appearance of the right lateral annulus with its more direct vertical motion. As expected, right PWD-MPI was only moderately [23] reliable (ICC 0.69), improved upon with TDI-MPI (ICC 0.83). These differences were notable yet did not reach statistical significance.

For the left side of the heart repeatability results for both PWD (0.80) and TDI (0.86) MPI measurement were good [23]. MAPSE results revealed only moderate repeatability with inter-observer ICC of 0.70 and average intra-observer ICC of 0.69. These differences did not reach statistical significance. As we have recently shown [27], TDI-MPI demonstrated very good intra-observer reliability (ICC 0.88–0.94) but low inter-observer ICCs (0.56), in part likely due to a high level of variation in waveform appearance. The poor correlation between TDI-MPI and PWD-MPI values, along with the variation in image appearance question the usefulness of TDI for MPI measurement. With intra- and inter-observer repeatability ICCs of 0.69 and 0.70 respectively MAPSE showed the poorest reliability within the left side evaluation, lower than PWD-MPI (ICC 0.80). Left TDI-MPI repeatability, similarly to the right-side assessment, revealed good intra-observer repeatability (ICC 0.86) although inter-observer repeatability could not be assessed. Given the same inconsistencies of appearance as seen in the right side of the heart, the automated PWD-MPI appears to be the most reliable for evaluation of left sided fetal cardiac function. For all three modalities evaluated for both sides of the heart, subgroup analysis did not find any significant differences or consistent trends comparing reliability between the different gestational age groups.

The main strength of this prospective study is that it focused on comparing three novel indices within both sides of the heart allowing a comprehensive multi-modality and global functional evaluation in a carefully selected cross-sectional study group of uncomplicated pregnancies.

This study had several limitations, including the research setting in which ultrasound scans were carried out. A dedicated research environment helps generating high acquisition rates for all cardiac function modalities offering more time to assess cardiac parameters. However, this could be at the fundamental expenses of generalizability of results to a clinical setting. Whilst there were notable differences in repeatability between the different modalities, these did not reach statistical significance for either side of the heart, which may have been overcome by a larger sample size. Use of only one ultrasound machine type may have been a limitation as our previous work has shown significant differences in TDI plots from different machines [27]. Finally, there is a lack of a gold standard fetal cardiac function modality for comparison. We acknowledge that other techniques for evaluating cardiac function are available (e.g. Doppler of AV valves and outflow tracts, cardiac biometry and indices of cardiac remodelling) and were not considered in this study.

Figure 3:

Left pulsed wave myocardial performance index (a, b), left tissue Doppler imaging myocardial performance index measurement (c, d) and Spatio-temporal image correlation Tricuspid annulus plane excursion (e, f) and mitral annulus plane excursion measurements (g, h).

Conclusions

While no single method is definitively superior to the others in both heart sides, our results on achievability and repeatability suggest TAPSE to be more suited to the right heart and PWD-MPI to the left heart. TAPSE provides a well-defined distinct M-mode waveform delineating the more longitudinal vertical movement of the tricuspid annulus, due to the longitudinal muscle fibre organisation of the dominant right ventricle. Automated PWD-MPI provides a verified standardised, automated technique consistently and reliably measuring the left heart inflow and outflow in a single cardiac plane.