CONTRIBUTION

What are the novel findings of this work?

Fetal atrioventricular plane displacement in the left and right ventricular walls and interventricular septum of the fetal heart (i.e. mitral, tricuspid and septal annular plane systolic excursion, respectively) can be analyzed automatically over several cardiac cycles using myocardial velocity traces obtained by color tissue Doppler imaging (cTDI).

What are the clinical implications of this work?

Automated analysis of cTDI cineloops of the four-chamber view of the fetal heart has the potential to simplify assessment of fetal atrioventricular plane displacement, enabling gathering of larger amounts of data. Such data could potentially be used in machine-learning models to facilitate prenatal assessment of different fetal disease states and evaluation of fetal risk at an individual level.

INTRODUCTION

Atrioventricular (AV) plane displacement (AVPD), often expressed as mitral annular plane systolic excursion (MAPSE), septal annular plane systolic excursion (SAPSE) and tricuspid annular plane systolic excursion (TAPSE), is a major contributor to cardiac pumping1, 2 and a measure of longitudinal cardiac function. AVPD is defined as the distance between the two positions where the AV plane is the farthest from the apex and closest to the apex during a cardiac cycle3. In healthy adults, examined using magnetic resonance imaging, AVPD accounts for 80% of the right-ventricular (RV) and 60% of the left-ventricular (LV) stroke volume4. Even in situations in which the absolute AVPD is decreased, the longitudinal contribution remains the main component of LV stroke volume5, 6. Longitudinal cardiac function, which is executed by longitudinal myocardial fibers, is considered to be the first affected in the event of hypoxia7, 8.

In fetuses, AVPD can be evaluated using ultrasound techniques, such as M-mode, tissue Doppler imaging (TDI) and speckle tracking3, 9. AVPD depends on body10 and cardiac11 size, and absolute values increase with gestational age (GA)12.

Using color TDI (cTDI), AVPD can be analyzed by integrating myocardial velocities. Through the application of an in-house automated algorithm, which has been used previously to analyze myocardial velocities and cardiac cycle time intervals in fetuses13, 14, automated measurements from several cardiac cycles can be obtained. Such an analysis allows for a quicker evaluation of large amounts of data that could potentially help in evaluating AVPD as a tool to detect cardiac dysfunction in fetuses with different disease states.

The primary aim of this study was to evaluate the feasibility of automated measurement of AVPD over several cardiac cycles from cTDI cineloops of the four-chamber view of the fetal heart. A comparison of cTDI-derived AVPD with AVPD measured using anatomic M-mode was also performed. The secondary objectives were, first, to establish reference ranges for AVPD during the second half of normal pregnancy, second, to assess AVPD in prolonged pregnancy in relation to adverse perinatal outcome and, third, to assess AVPD in fetuses with suspicion of intrauterine growth restriction (IUGR).

METHODS

This was a cross-sectional study conducted at the Center for Fetal Medicine, Karolinska University Hospital, Stockholm, Sweden, between September 2009 and February 2017. Ethical approval was obtained from the Stockholm Regional Ethics Committee (DNr 2009/1617-31/2, 2012/895-31/4 and 2017/539-32). All women gave written informed consent to participate.

The first cohort, used for the development of reference ranges for AVPD in the second half of pregnancy, consisted of women with an uncomplicated singleton pregnancy between 18 and 42 weeks of gestation (n = 201). Exclusion criteria were conception after in-vitro fertilization/intracytoplasmic sperm injection treatment, maternal complications at inclusion, such as pre-eclampsia, chronic hypertension or diabetes, and fetal chromosomal or major structural abnormalities discovered during pregnancy or postnatally. Pregnancies were dated by measuring the biparietal diameter in the second trimester15. Each fetus was included only once. Reference ranges for myocardial velocities and cardiac cycle time intervals based on the same population have been published previously14.

The second cohort, used to assess AVPD in prolonged pregnancy in relation to adverse perinatal outcome, comprised women with an uncomplicated pregnancy who attended a routine appointment at ≥ 41 + 0 weeks of gestation (n = 107) and underwent an ultrasound examination between 41 + 0 and 41 + 5 weeks. A composite adverse perinatal outcome was defined as the presence of at least one of the following: intrapartum fetal scalp blood lactate > 4.8 mmol/L, umbilical cord arterial pH < 7.15, 5-min Apgar score < 7 or 5-min Apgar score for muscle tone < 2. Data regarding the feasibility of automated analysis of fetal myocardial velocity measurements obtained by cTDI have been published previously using this cohort16.

The third cohort included, as part of a pilot study, women with a singleton fetus with a suspicion of IUGR, defined as an estimated fetal weight (EFW) < 2.5th centile or an EFW < 10th centile and umbilical artery pulsatility index > 97.5th centile. The definition used is based on Swedish reference ranges for estimation of fetal weight and umbilical artery cut-off values17, 18.

Fetal echocardiography using cTDI was performed by an experienced ultrasonographer (K.F.-W.) using a GE Vivid-i ultrasound imaging system, equipped with a 3S-RS 1.9–3.8-MHz phased-array transducer (GE Vingmed, Horten, Norway) or a Vivid S6 ultrasound imaging system with a M4S-RS 1.9–4.1-MHz phased-array transducer (GE CV Ultrasound, Haifa, Israel). An apical or basal four-chamber view of the fetal heart was acquired and cineloops of consecutive cardiac cycles were recorded using cTDI. The insonation angle was kept as close as possible to the long axis of the heart (and always < 30°) and the image was adjusted to obtain as high a frame rate as possible. Offline analysis was performed using EchoPAC version 201 (GE Vingmed). The region of interest (ROI) was adjusted in height and width according to GA, as in previous research13, and placed at the AV plane in the LV and RV walls and the interventricular septum (IVS). All ROIs were placed by one operator (L.H.). Myocardial velocity traces were subsequently exported from EchoPAC to MATLAB (R2019b; MathWorks, Natick, MA, USA) and then analyzed using an automated software tool developed in-house. Cardiac cycle time intervals were defined by the software. The program in MATLAB, i.e. the automated analysis described, has several user-defined functions and some MATLAB built-in functions2, 19. The myocardial velocity data were integrated and AVPD was obtained automatically and defined as the maximum distance covered by the AV plane during the cardiac cycle (Figure 1). AVPD in the LV, IVS and RV were called MAPSE, SAPSE and TAPSE, respectively, and according to nomenclature used previously in fetuses. The analysis is a fully automated procedure and no manual marking of the traces is needed once the velocity traces are exported to MATLAB. The average of all available cardiac cycles was also calculated.

Automated analysis of a myocardial velocity trace from the right ventricular wall (a) and a trace showing the tricuspid annular peak systolic excursion (TAPSE) (b), obtained by color tissue Doppler imaging in a normal 34-week fetus. The colors indicate the different phases in the cardiac cycle, as defined by the automated algorithm: atrial contraction (), pre-ejection (), ventricular ejection (), post-ejection (), rapid ventricular filling (), slow ventricular filling (). The blue and red asterisks in (b) indicate the maximum amplitude of TAPSE.

The fetal cardiac size was measured as the longitudinal diameter (apex to base) in early systole, right after the closure of the AV valves in diastole. Calipers were placed from the midpoint of the atria at the outer border to the outer border of the apex. The cardiac size, i.e. the longitudinal diameter of the heart20, was used to normalize AVPD measurements.

In a subset of 30 fetuses from the first cohort (i.e. the cohort used for development of the reference ranges), AVPD was measured additionally using anatomic M-mode echocardiography for comparison with the cTDI-derived AVPD measurements obtained using the automated technique. The M-mode cursor was placed through the lateral leaflets of the AV valves and the IVS to obtain an alignment with the IVS < 30°. The total AVPD was measured according to previously described techniques21, 22. The average of three measurements was calculated.

The inter- and intraobserver variability of the acquisition of cTDI recordings and offline analysis was evaluated using a Vivid S6 ultrasound imaging system with a M4S-RS 1.9–4.1-MHz phased-array transducer (GE CV Ultrasound). The interobserver variability was evaluated by two ultrasonographers (K.F.-W. and L.H.) who examined 25 fetuses, between 19 and 41 weeks of gestation, between December 2016 and February 2017. Subsequently, 22 of these fetuses were examined again by the first ultrasonographer (K.F.-W.) approximately 10 min later, and the intraobserver variability for the entire procedure was evaluated by comparing the two recordings. This means that the calculated coefficients of variation (CV) also reflect the variability associated with the change in fetal physiological state and position, as well as the resulting technical challenges during the examination period, not just the operator-related measurement variability. ROIs were placed by one operator (L.H.) and all generated myocardial velocity traces were analyzed by the automated algorithm.

Statistical analysis

Data analysis was performed using IBM SPSS Statistics for Windows, version 25.0 (IBM Corp., Armonk, NY, USA) and MATLAB (MathWorks). Continuous variables are presented as mean ± SD or median (interquartile range), as appropriate. Categorical variables are presented as n (%). The statistical approach described by Royston and Wright23 was used to create reference intervals for MAPSE, SAPSE and TAPSE. Box–Cox power transformation was used to decide the appropriate type of transformation. The best-fitting fractional polynomials were chosen from a list of 44 regression models based on R2 value, using NCSS 2020 Statistical Software (NCSS LLC, Kaysville, UT, USA) to construct mean curves of each dependent variable in relation to GA, expressed in exact weeks (decimal days). The normality of the distribution of the Z-scores was checked using the Shapiro–Wilk test. Mean, SD and CI curves as a function of GA were calculated and plotted. The equations of the polynomial regression were used to calculate the estimated mean, 5th and 95th centiles for the corresponding GA as fitted mean ± 1.645 SD. CIs were calculated for the fitted mean, 5th and 95th centiles. The same procedure was then performed for values normalized for cardiac size.

The MAPSE, SAPSE and TAPSE of the prolonged pregnancies with an adverse neonatal outcome were plotted on the reference ranges of the normal population. The mean Z-scores of AVPD in the cohort of prolonged pregnancies were compared between cases with a normal outcome and those with an adverse outcome using Mann–Whitney U-test.

The AVPD measurements in the IUGR cohort were plotted on the graphs of the reference ranges and the mean Z-scores of AVPD were compared between the IUGR group and the normal population using Mann–Whitney U-test.

The correlation between AVPD measurements obtained using cTDI and those obtained by anatomic M-mode was evaluated using Spearman's correlation, and agreement between the two measurement methods was tested for bias and precision using Bland–Altman analysis of log-transformed data24-26. We defined precision by the 95% limits of agreement (LoA) (± 1.96 SD) and bias as the difference between the geometric mean of all individual ratios between pairs of values obtained using the two methods and the line of equality. The statistical significance level was set to P < 0.05. To assess intra- and interobserver variability, CVs were calculated using the root mean square method27.

RESULTS

Baseline characteristics, ultrasound data and pregnancy outcomes of the normal group (n = 201), prolonged pregnancies (n = 107) and fetuses with a suspicion of IUGR (n = 35) are presented in Table 1.

Table 1. Maternal baseline characteristics, ultrasound data and pregnancy outcome in 201 normal pregnancies between 18 and 42 weeks' gestation, 107 pregnancies at ≥ 41 weeks' gestation and 35 pregnancies with a suspicion of intrauterine growth restriction (IUGR) Variable Normal cohort (n = 201) Prolonged pregnancy (n = 107) Suspected IUGR (n = 35) Maternal data Age (years) 30.1 ± 5.2 31.0 ± 5.2 30.8 ± 5.3 BMI (kg/m2) 23.7 ± 3.9 24.9 ± 5.2 25.1 ± 5.5 Spontaneous pregnancy 201 (100) 100 (93.5) 31 (88.6) Nulliparous 81 (40.3) 45 (42.1) 17 (48.6) Smoker 12 (6.0) 2 (1.9) 6 (17.1) Ultrasound data GA at scan/inclusion (weeks) 30.4 ± 7.2 41.1 ± 0.2 33.2 ± 3.4 UA-PI > 97.5th centile 0 (0) 4 (3.7) 14 (40.0) Apical position of fetal heart 107 (53.2) 32 (29.9) 14 (40.0) Frame rate 208.3 ± 6.8 206.3 ± 15.5 208.1 ± 10.4 Pregnancy outcome* Pre-eclampsia 4 (2.0) 0 (0) 11 (31.4) GA at delivery (weeks) 40.0 ± 1.4 41.7 ± 0.4 35.6 ± 3.8 Post-term delivery ≥ 42 + 0 weeks 19 (9.6) 35 (32.7) 0 (0) Preterm delivery < 37 weeks 4 (2.0) 0 (0) 19 (54.3) Mode of delivery Normal vaginal 147 (74.2) 85 (79.4) 18 (51.4) Vacuum extraction 22 (11.1) 9 (8.4) 0 (0) Cesarean section 29 (14.6) 13 (12.1) 17 (48.6) Elective NA 3 (2.8) 1 (2.9) Emergency NA 10 (9.3) 16 (45.7) Birth weight (g) 3576 ± 501 3840 ± 423 1973 ± 632 Female neonate 96 (48.5) 55 (51.4) 21 (60.0) Cord arterial pH† 7.25 ± 0.09 7.22 ± 0.09 7.29 ± 0.09 Cord venous pH‡ 7.32 ± 0.07 7.31 ± 0.07 7.38 ± 0.06 5-min Apgar score < 7 0 (0) 1 (0.9) 2 (5.7) Data are given as mean ± SD or n (%). * Pregnancy outcome data were missing for three women in the control group who delivered in other hospitals. Data available for: † 138, 83 and 21 cases in the normal, prolonged-pregnancy and IUGR cohorts, respectively; ‡ 49, 98 and six cases in the normal, prolonged-pregnancy and IUGR cohorts, respectively. BMI, body mass index; GA, gestational age; NA, not available; PI, pulsatility index; UA, umbilical artery. Normal reference ranges for cTDI-derived AVPD

Initially, 202 pregnant women were enrolled in this cohort. After the exclusion of one woman whose fetus was diagnosed with muscular ventricular septal defect and bicuspid aortic valve, the cohort consisted of 201 women. In total, 603 fetal myocardial velocity traces (201 each from LV, IVS and RV) were available for analysis and subsequent integration to obtain AVPD data. One RV trace was excluded due to severe artifacts and two IVS traces could not be analyzed. The mean ± SD number of cardiac cycles analyzed were 10.2 ± 2.6 in the LV, 10.3 ± 2.7 in the IVS and 10.3 ± 2.6 in the RV.

The interobserver CVs for the automated measurement of AVPD on cTDI cineloop recordings of the four-chamber view of the fetal heart obtained consecutively by two sonographers in 25 fetuses were 28.1%, 17.7% and 15.3% for MAPSE, SAPSE and TAPSE, respectively. The intraobserver CVs for the automated measurement of AVPD on cineloop recordings obtained by a single sonographer approximately 10 min apart in 22 fetuses were 33.5%, 15.0% and 17.9% for MAPSE, SAPSE and TAPSE, respectively. The measurement variability of automated analysis on the same cineloop recording was zero on repeated evaluation.

The best model for all variables was a quadratic polynomial fit. MAPSE was transformed logarithmically, whereas SAPSE and TAPSE did not require any transformation. MAPSE, SAPSE and TAPSE all increased with GA (Figure 2 and Tables 2–4). The regression equations are displayed in Table 5. The fitted mean was highest in TAPSE throughout the second half of gestation, followed by SAPSE and MAPSE. Reference ranges with the corresponding 95% CIs calculated for the fitted mean and 5th and 95th centiles are displayed in Figure 2. A Z-score and centile calculator for MAPSE, SAPSE and TAPSE obtained by cTDI and analyzed automatically is provided in Appendix S1.

(a) Scatterplots of mitral annular plane systolic excursion (MAPSE), septal annular plane systolic excursion (SAPSE) and tricuspid annular plane systolic excursion (TAPSE) obtained by color tissue Doppler imaging and analyzed automatically, in 201 low-risk pregnancies (), 35 fetuses with a suspicion of intrauterine growth restriction (IUGR) () and 21 fetuses at ≥ 41 + 0 weeks of gestation that had an adverse outcome (), plotted against gestational age. The fitted mean and 5th and 95th centiles, with corresponding 95% CIs (), of the low-risk cohort are shown. (b) Corresponding Z-scores for MAPSE, SAPSE and TAPSE. , mean in fetuses with suspicion of IUGR; , mean in prolonged pregnancies with adverse outcome. , ± 1.645 SD. Table 2. Measurements of fetal mitral annular plane systolic excursion (in mm), obtained by color tissue Doppler imaging and analyzed automatically, according to gestational age (GA) Centile GA (weeks) n 2.5th 5th 10th 50th 90th 95th 97.5th 18 15 1.11 1.19 1.29 1.74 2.34 2.55 2.74 19 5 1.16 1.25 1.37 1.86 2.53 2.76 2.98 20 4 1.22 1.31 1.44 1.98 2.73 2.99 3.23 21 8 1.27 1.38 1.51 2.11 2.93 3.22 3.50 22 4 1.32 1.43 1.58 2.23 3.14 3.46 3.77 23 8 1.37 1.49 1.65 2.35 3.35 3.71 4.05 24 6 1.41 1.54 1.71 2.47 3.57 3.96 4.33 25 12 1.45 1.59 1.77 2.59 3.78 4.21 4.62 26 6 1.49 1.64 1.83 2.71 4.00 4.47 4.91 27 11 1.52 1.68 1.88 2.82 4.21 4.72 5.21 28 6 1.55 1.72 1.93 2.92 4.42 4.97 5.50 29 8 1.58 1.75 1.97 3.02 4.63 5.22 5.79 30 10 1.60 1.78 2.01 3.11 4.82 5.46 6.08 31 10 1.61 1.80 2.04 3.20 5.01 5.69 6.36 32 9 1.62 1.81 2.07 3.28 5.19 5.92 6.63 33 8 1.62 1.82 2.08 3.34 5.36 6.13 6.88 34 13 1.62 1.83 2.09 3.40 5.52 6.33 7.13 35 5 1.61 1.82 2.10 3.45 5.66 6.51 7.36 36 5 1.60 1.81 2.10 3.48 5.78 6.68 7.57 37 8 1.58 1.80 2.09 3.51 5.89 6.83 7.76 38 8 1.56 1.78 2.07 3.52 5.98 6.96 7.93 39 4 1.53 1.75 2.05 3.52 6.06 7.06 8.07 40 4 1.50 1.72 2.02 3.51 6.11 7.15 8.19 41 15 1.47 1.69 1.98 3.49 6.14 7.21 8.29 42 9 1.43 1.64 1.94 3.45 6.16 7.25 8.36 Table 3. Measurements of fetal septal annular plane systolic excursion (in mm), obtained by color tissue Doppler imaging and analyzed automatically, according to gestational age (GA) Centile GA (weeks) n 2.5th 5th 10th 50th 90th 95th 97.5th 18 15 1.09 1.26 1.44 2.10 2.77 2.95 3.12 19