First trimester two- and four-dimensional cardiac scan: intra- and interobserver agreement, comparison between methods and benefits of color Doppler technique

Authors

  • S. Tudorache,

    Corresponding author
    1. Prenatal Diagnostic Unit, Emergency University Hospital, Craiova, Dolj, Romania
    2. University of Medicine and Pharmacy Craiova
    • Correspondence to: Dr S. Tudorache, University of Medicine and Pharmacy Craiova, Department of Obstetrics and Gynecology, 2-4 Petru Rares, Craiova, Dolj, 200349, Romania; Emergency University County Hospital, First Clinic of Obstetrics and Gynecology, 1 Tabaci, Craiova, Dolj, 200642, Romania (e-mail: stefania.tudorache@gmail.com)

    Search for more papers by this author
  • M. Cara,

    1. University of Medicine and Pharmacy Craiova
    2. Public Health Department Craiova, Craiova, Dolj, Romania
    Search for more papers by this author
  • D. G. Iliescu,

    1. Prenatal Diagnostic Unit, Emergency University Hospital, Craiova, Dolj, Romania
    2. University of Medicine and Pharmacy Craiova
    Search for more papers by this author
  • L. Novac,

    1. University of Medicine and Pharmacy Craiova
    Search for more papers by this author
  • N. Cernea

    1. Prenatal Diagnostic Unit, Emergency University Hospital, Craiova, Dolj, Romania
    2. University of Medicine and Pharmacy Craiova
    Search for more papers by this author

ABSTRACT

Objective

To evaluate intra- and interobserver agreement for first-trimester fetal cardiac structural assessment, using two-dimensional (2D) ultrasound (2D-US) and 4D-US (4D spatiotemporal image correlation (STIC) technology), to compare the methods and to assess the advantages of adding color Doppler to each technique.

Methods

Digital videoclips (B-mode and color Doppler) and 4D-STIC volumes (gray-scale and color Doppler) from 632 pregnancies with normal fetal hearts were acquired and stored at the time of detailed first-trimester ultrasound examination. Later analysis on a randomized sample of 100 cases was performed, targeting 11 cardiac structures and features. We compared visualization of fetal heart parameters using 2D-US vs 4D-US and gray-scale vs color Doppler imaging.

Results

STIC volumes were considered satisfactory (adequate visualization of at least 8/11 parameters) in 78% of cases and 2D-US acquisitions in 89% of cases. The intra- and interobserver agreement was good for both 2D and 4D methods (kappa > 0.6), and the percentage overall agreement was very high using both methods (95%). 2D- and 4D-US identification of the fetal cardiac parameters did not differ significantly. The differences between gray-scale and color Doppler imaging were statistically significant in identifying similar key cardiac parameters, for both 2D- and 4D-US (P < 0.05).

Conclusion

Both 2D and 4D methods for assessing first-trimester heart parameters are feasible and repeatable within and between observers. Color Doppler adds valuable information to both methods. Copyright © 2013 ISUOG. Published by John Wiley & Sons Ltd.

INTRODUCTION

In recent years we have witnessed the introduction of first-trimester screening for aneuploidy and fetal extracardiac anomalies[1, 2], while first-trimester cardiac assessment is not routine, still being regarded as an advanced technique involving high-resource settings and skilled operators[3]. It is, however, possible to perform a first-trimester cardiac scan using two-dimensional ultrasound (2D-US)[4-7], and we believe that the heart should be looked at and ultimately screened at this time. There are several reasons for this: at this gestational age the heart has almost completed its development and most congenital heart diseases (CHD) are already present and thus potentially detectable[4-6]; CHD are usually not associated with known risk factors[4-6, 8, 9]; they have a higher prevalence than do other fetal defects[5, 7]; they are the leading cause of infant mortality[9, 10]; those detectable in the first trimester are generally more severe[2, 4, 6, 7, 11, 12]; and first-trimester termination of pregnancy is safer and easier to perform compared with later in pregnancy[7].

First-trimester imaging should be designed for screening, not for refining diagnosis. The potential advantages of performing first-trimester fetal cardiac evaluation using conventional 2D-US, which is probably capable of accurately detecting major CHD[13], are its simplicity and low cost, making it more appropriate for large-scale application. The main disadvantage is its operator-dependency. 4D-US (i.e. spatiotemporal image correlation (STIC)) has the advantage of being less operator-dependent and permitting offline expert reassessment of volumes[14-19], and has been proposed as a feasible and reliable alternative[16, 19], but the technique requires high-quality systems and probes and implies a need for telemedicine facilities. Extending the available data regarding the intra- and interobserver agreement of these two imaging techniques for cardiac assessment in the first trimester would provide valuable clinical information about their reliability, while visualization rates of the various cardiac ultrasound parameters using the different methods of assessment would indicate their comparative value in confirming normal heart anatomy.

Our objectives were to investigate intra- and interobserver agreement for 2D-US and 4D-US cardiac structural assessment in the first trimester, to compare these two methods and to assess the advantages of using color Doppler in addition to gray-scale imaging for each. Our hypotheses were that 2D-US and 4D-US examination using traditional planes would prove to have comparable reproducibility in normal heart evaluation and that addition of color Doppler would better demonstrate the searched ultrasound parameters.

METHODS

Study design and setting

This was an observational study of singleton pregnancies in a low-risk population undergoing first-trimester genetic and detailed anomaly scan at 12–14 gestational weeks in the prenatal diagnostic unit of County Emergency University Hospital Craiova, Romania, from January 2011 to May 2012.

Participants

There are around 3000 births/year at our University Hospital. Due to late booking, not all are examined at the prenatal diagnostic unit in the first trimester. We considered for eligibility all 1605 consecutive low-risk pregnancies that booked before 13 + 6 weeks and we declared as eligible the 1464 cases that fulfilled our inclusion criteria: gestational age between 12 + 0 and 13 + 6 weeks (crown–rump length 54–78 mm), normal maternal body mass index (BMI, 18.5–25 kg/m[2]), low risk for CHD according to history and nuchal translucency thickness and absence of structural abnormalities or chromosomal markers at first-trimester scan. We chose this gestational age in accordance with the generally accepted timing of the first-trimester genetic scan, but we decided to limit our investigation to fetuses above 12 gestational weeks due to the likelihood of suboptimal images being obtained below this age.

Using the hospital coding system, and the patient's (hard copy and electronic) files, we subsequently identified postpartum 632 cases with normal second-trimester echocardiogram and known normal perinatal outcome (neonatal clinic and sonographic heart exam at discharge, term birth). From these cases we selected for entry into the final statistical analysis a sufficient number of cases (n = 100) for demonstration of the objectives, using a random number generator. Their files were exported to personal computers using the ‘anonymize personal data’ and ‘anonymize clinical data’ options of the 4D View Sonoview (GE Medical Systems, Zipf, Austria) software package.

Scans and analysis were performed by two obstetrician sonographers (S.T. (Observer 1) and D.G.I. (Observer 2)), each with more than 10 years' experience in prenatal diagnosis and first-trimester assessment and specialized in fetal echocardiography.

The study protocol was approved by the university ethics committee and informed consent was obtained from all participants at inclusion prior to scanning.

Variables

For both 2D-US and 4D-US methods, using both gray-scale and color Doppler techniques, the outcome of each scan was assessed according to satisfactory or unsatisfactory visualization of a normal view for a set of 11 fetal cardiac ultrasound parameters used for the confirmation of normal heart anatomy.

We considered that the results of our study (inter- and intraobserver agreement, visualization rates) would not be influenced significantly by the confounders that can influence image quality in the clinical setting, such as placental position (anterior/posterior), patient-specific acoustic characteristics of body tissue, subjectively assessed amniotic fluid volume, bowel filling, uterus position (incomplete anteversion or anteflexion), scarred uterus or scarred abdominal wall. Cases were not, therefore, subanalyzed on the basis of these characteristics.

Data acquisition

In all 1464 cases meeting the first-trimester inclusion criteria, ultrasound examinations were performed transabdominally by Observer 1, using a Voluson 730 Expert series and an E8 (GE Medical Systems) ultrasound machine, equipped with a 4–8-MHz curvilinear transducer. All scans were performed using speckle reduction imaging and CrossXBeam CRI (contrast resonance imaging). When using color Doppler the mechanical and thermal indices were kept as low as possible (ALARA principle) and safety guidelines were followed[20-22].

In the first stage, the 4D volumes and 2D videoclips of the heart were obtained by Observer 1 for all eligible cases. This was the only stage in which the observer was not blinded to the patient's clinical information.

In our experience, an oblique lateral insonation from the right part of the thorax is optimal for visualization of the three planes essential for first-trimester fetal cardiac evaluation: the four-chamber view (4-CV) plane, five-chamber view plane and three vessels and trachea view plane. Thus, the protocol for heart evaluation was as follows: the operator waited and focused on other segments of fetal anatomy until the fetus offered a lateral view from the right side of the thorax (or from the right shoulder) or an apical view of the heart; the angle beam was narrowed to 15–20°; high-definition zoom was applied, until only the thorax was visible, occupying approximately 80% of the screen; the operator then turned on the color flow gate (Videoclips S1 and S2). This protocol was adopted due to the ongoing concerns of all ultrasound societies regarding the level of first-trimester fetal exposure to ultrasound energy[21, 22]; its purpose was to add very little to the amount of ultrasound energy to which the fetus was exposed.

The 4D-STIC volumes were acquired first, using both gray-scale and color Doppler modes. We chose as the standard acquisition plane that of the five-chamber view (which in the first trimester is the same as the ‘crossing plane’), using a lateral (or an apical) approach. The acquisition time was 12.5 s and the acquisition angle 15–20°. The color Doppler maximal velocity setting was adjusted such that the great arteries appeared equal in size, were homogeneous in color and did not demonstrate aliasing. At least two volumes (apical or lateral view, one gray-scale and one color flow mapping) per case were stored on the hard drive of the ultrasound machine with the STIC software for later offline analysis with 4D View Sonoview (GE Medical Systems). The observer was allowed to prolong the usual nuchal translucency scan time by a maximum of 10 min for the cardiac study.

The 2D-US data were then obtained and analyzed by Observer 1, using both gray-scale and color Doppler modes. Cardiac sweeps were performed, moving from the abdominal plane to the transverse aortic arch (Figure 1). At least two 2D digital videoclips (one gray-scale and one color flow mapping) per case were stored on the hard drive of the ultrasound machine, for later analysis.

Figure 1.

Protocol used for assessment of cardiac structures, showing each of the four planes in lateral view from the right shoulder (a,c,e,g) and in apical view (b,d,f,h), on both gray-scale (left of each pair of images) and color Doppler (on right) imaging. (a,b) Transverse abdominal plane: gray-scale imaging shows relation of heart and stomach, and abdominal aorta (with option of identifying inferior vena caval location). (c,d) Four-chamber view plane: gray-scale imaging shows four-chambered heart, crux cordis and one pulmonary vein entering left atrium; color Doppler imaging shows equal atrioventricular flow and no flow between ventricles (intact ventricular septum). (e,f) Five-chamber view (crossing plane): gray-scale imaging shows emergence of left ventricular outflow tract and septoaortic continuity; color Doppler imaging shows aortic flow and crossover of the great arteries. (g,h) Three vessels and trachea view plane: gray-scale imaging shows confluence of arterial arches on left of spine; color Doppler imaging shows confluence of arterial arches and normal direction and equal flow in both arches.

During the 2D online examination, Observer 1 filled in a form which included 11 cardiac anatomical ultrasound parameters (six structures on gray-scale imaging and five features on color flow mapping) (Figure 1), defining each as being satisfactory or non-satisfactory for evaluation, i.e. noting ‘yes’ and ‘no’ (or ‘+’ and ‘–’), accordingly. On gray-scale imaging we searched for: (1) abdominal situs, in the transverse abdominal plane; (2) four-chambered heart (i.e. chambers numbering four), (3) intact crux and (4) at least one pulmonary vein entering the left atrium, all in the four-chamber view plane; (5) left ventricular outflow tract (LVOT) and septoaortic continuity (SAC) in the five-chamber view plane; and (6) confluence of the arterial arches on the left of the spine, in the three vessels and trachea view plane. On color flow imaging we searched for: (1) equal atrioventricular flow and (2) no flow between ventricles (i.e. intact ventricular septum), in the four-chamber view plane; (3) aortic flow and (4) crossover of the great arteries, in the five-chamber view plane; and (5) confluence of the arterial arches with normal direction and equal flow in both arches, in the three vessels and trachea view plane[8, 10, 23, 24] (Figure 1).

In the second stage, following delivery of all pregnancies and identification of a random sample of 100 cases from the ones that remained eligible following exclusions, Observer 2, blinded to the findings of Observer 1, analyzed the 2D digital videoclips and filled in a similar form.

In the third stage, 2 weeks later, Observer 1, blinded to the previous findings, assessed the self-acquired 4D volumes over a period of 25 min using 4D View, searching for the same 11 parameters established in the 2D examination and filling in an offline analysis form for volumes. The assessment included sweeping from the initial acquisition plane, in the caudal or cranial direction, zooming, applying tomographic ultrasound imaging, slowing or freezing the cardiac motion to obtain views and structures, and use of swing technique or fast echo algorithm[25, 26] (Videoclip S3).

In the fourth and final stage, a further 2 weeks later, Observer 2, blinded to previous evaluations, assessed the stored 4D volumes and filled in the form in the same way as had Observer 1; this was at least 4 weeks after having performed the 2D analysis in stage 2.

The 2D digital videoclips and 4D-STIC acquisitions were declared ‘satisfactory’ overall when at least eight of the 11 ultrasound parameters were visualized ‘satisfactorily’ and ‘insufficient’ if visualization of four or more parameters failed.

In order to assess intraobserver agreement, 2D-US and 4D-US datasets were reanalyzed randomly, by each observer, 1 month after the end of the study.

Results of the patients' second-trimester evaluations and details of pregnancy outcome were obtained from the hospital database.

Study size

The sample size was chosen a priori, based on the goodness-of-fit formula of Donner and Eliasziw, considering that we wanted to detect a statistically significant kappa value of 0.9 (P < 0.5) on a dichotomous variable with 80% power, and assuming the null hypothesis value of kappa to be 0.5 and a proportion of positive ratings of 0.9[27].

Bias

Due to the fact that we used only normal cases, studied by the same standardized technique, recall bias should have been very low. Still, we addressed the possible bias due to memory by establishing a minimum 4-week period between repeat examinations, and blinding the observers to their own and each other's previous results. Images were interpreted independently by the observers, but under similar conditions (level of light in room and using same hardware). Different random ordering of patients on each rating occasion and for each observer was used for the same purpose.

Chance effects were minimized by ensuring the study was adequately powered.

The main source of potential bias, in our opinion, would be the operator's awareness of a case's specific features. This would only have affected stage 1, and we addressed this issue by having Observer 1 collect the 4D-US datasets before the 2D-US digital videoclips. An unavoidable potential source of bias was that the two observers had spent the last 10 years working closely together in the same sonography unit. This factor may have produced better agreement levels than would have been obtained if the study had included observers from different institutions.

Statistical analysis

The primary outcomes in this study were observer variation quantified by the kappa statistic for nominal data (dichotomous yes/no answers), the comparison between two different methods of scanning (2D-US and 4D-US) and the comparison between gray-scale and color Doppler techniques.

The kappa value indicates the observed agreement which is above and beyond that due to chance, and is calculated as: math formula, where the observed agreement is the proportion of cases on which the observers agree, defined as the number of occasions of complete agreement divided by the total number of occasions, and the expected agreement is the probability that two people will provide by chance the same response to a question for any given patient. A kappa value of 1.00 means perfect agreement while a value of 0.00 means agreement equal to that expected from chance alone.

Cohen's kappa coefficient was calculated to evaluate intra- and interobserver agreement[27]. A kappa value of 0.21–0.40 was considered to show low agreement, 0.41–0.60 moderate agreement, 0.61–0.80 good agreement and 0.81–1.0 excellent agreement[28].

McNemar's test for paired proportions was used for the comparison of visualization of fetal heart structures between 2D and 4D ultrasound. McNemar's test was also used to compare gray-scale and color Doppler modes in the three traditional planes by assessing image quality (i.e. visualization) in pairs of similar targeted items, one in each mode: in the four-chamber view plane we compared visualization of the four-chambered heart (gray-scale) with that of equal atrioventricular flow (color Doppler); in the five-chamber view plane we compared visualization of the LVOT and SAC (gray-scale) to that of aortic flow (color Doppler); and in the three vessels and trachea view plane we compared visualization of the confluence of the arterial arches on the left of the spine (gray-scale) to that of equal flow in both arches (‘V sign’ on color Doppler). P < 0.05 was considered statistically significant.

Statistical analyses were performed using the statistical software IBM SPSS Statistics for Windows, Version 19.0 (IBM Corp., Armonk, NY, USA). For the reasons explained above, we did not use statistical methods to control for confounding.

RESULTS

Among the 100 randomized cases in the final analysis, the median maternal age was 32 (range, 20–43) years, median BMI was 22.8 kg/m2, median gestational age was 12.4 weeks and median number of 4D volumes per patient was three (range, two to five).

In the first stage, the fetal heart study was complete (with all 11 parameters demonstrated by 2D-US acquisition) in 52 cases and considered satisfactory for interpretation (with eight or more of the 11 parameters visualized satisfactorily) in 89 cases, and all color Doppler parameters were obtained in 86 cases.

There were 329 4D-STIC volumes included in the final analysis from the 100 randomized cases. STIC volumes were declared satisfactory for post-processing (at least 8/11 parameters) by both operators in 78 fetuses. In 77 cases, all color Doppler parameters were obtained. In 47 cases, at least one of the intended 4D acquisitions was stopped during the first stage, due to sudden fetal movement (101 volumes in total).

The visualization rate (number of times a given parameter was observed satisfactorily, expressed as a percentage of the total number of normal hearts (n = 100)) of every cardiac parameter by each observer is represented graphically in Figure 2. The visualization rate on gray-scale imaging for both 2D and 4D methods was good (≥ 80%) for most cardiac structures, the exceptions being the visualization rates of at least one pulmonary vein entering the left atrium, the crux and confluence of the arterial arches on the left of the spine in the three vessels and trachea view (Figure 2). On color flow mapping the visualization rate was excellent (≥ 85%) for all parameters.

Figure 2.

Visualization rate of cardiac structures on two-dimensional (2D) and 4D ultrasound by Observers 1 and 2 (image, 2D Observer 1; image, 2D Observer 2; image, 4D Observer 1; image, 4D Observer 2). Gray-scale imaging: abdominal situs, four-chambered heart, intact crux, at least one pulmonary vein entering left atrium (≥ 1 PV in LA), left ventricular outflow tract (LVOT) and septoaortic continuity (SAC), confluence of arterial arches on left of spine (Confl. arches); color Doppler: equal atrioventricular (AV) flow, intact ventricular septum (IVS), aortic flow (Ao flow), crossover of great arteries (GA), equal flow in both arches.

Kappa values for the intraobserver and interobserver repeatability analysis are presented in Tables 1 and 2, respectively. For all 11 cardiac features, for both observers, there was good or excellent intraobserver agreement (kappa > 0.6, Table 1). The overall intraobserver agreement was  > 95% for all cardiac features, for both observers and for both methods. Using 2D-US, Observer 1 showed good intraobserver agreement (kappa, 0.61–0.80) for four-chambered heart, intact ventricular septum and crossover of the great arteries, and excellent agreement for the other eight cardiac features (Table 1). Observer 2 showed excellent intraobserver agreement for all 11 cardiac parameters. Using 4D-US, both Observers 1 and 2 showed good intraobserver agreement for only one feature, abdominal situs, and for the remaining 10 features they each showed excellent intraobserver agreement.

Table 1. Intraobserver agreement for evaluation of fetal cardiac parameters, using two-dimensional (2D) and 4D ultrasound (US)
Cardiac parameter2D-US4D-US
Observer 1Observer 2Observer 1Observer 2
Po (%)Ppos (%)Pneg (%)k (95%CI)Po (%)Ppos (%)Pneg (%)k (95%CI)Po (%)Ppos (%)Pneg (%)k (95%CI)Po (%)Ppos (%)Pneg (%)k (95%CI)
  1. 3VT, three vessels and trachea; Ao, aortic; AV, atrioventricular; Confl. arches, confluence of arterial arches on left of spine; GA, great arteries; IVS, intact ventricular septum; k, kappa; LA, left atrium; LVOT, left ventricular outflow tract; P0, overall proportion of agreement (number of readings (both positive and negative) on which each observer agrees with himself, divided by total number of all readings performed (where total number = 100)); Ppos, positive agreement (number of positive findings on which each observer agrees with himself divided by total number of positive readings of the same observer); Pneg, negative agreement (number of negative findings on which each observer agrees with himself divided by total number of negative readings of the same observer); PV, pulmonary vein; SAC, septoaortic continuity.

Gray-scale                
Abdominal situs1001001001001001009798770.759798770.75
            (0.48–1.00)   (0.48–1.00)
Four-chambered heart9899750.749999860.859899860.859899860.85
    (0.38–1.00)   (0.56–1.00)   (0.64–1.00)   (0.64–1.00)
Intact crux9798960.941001001001.009898980.969999990.98
    (0.87–1.00)   (1.00–1.00)   (0.90–1.00)   (0.94–1.00)
At least one PV entering LA9797970.949999990.989999990.989999990.98
    (0.87–1.00)   (0.94–1.00)   (0.94–1.00)   (0.94–1.00)
LVOT and SAC9597850.829698880.869899950.949999980.97
    (0.67–0.97)   (0.72–0.99)   (0.86–1.00)   (0.91–1.00)
Confl. arches9595950.901001001009999990.989898980.96
    (0.81–0.99)       (0.94–1.00)   (0.91–1.00)
Color Doppler                
Equal AV flow9798860.849999960.959899900.899999960.95
    (0.66–1.00)   (0.87–1.00)   (0.74–1.00)   (0.86–1.00)
IVS9798770.759999940.949999950.949999960.95
    (0.48–1.00)   (0.81–1.00)   (0.83–1.00)   (0.87–1.00)
Ao flow9899860.859899860.859999950.94100100100
    (0.64–1.00)   (0.64–1.00)  (0.83–1.00)(0.74–1.00)    
Crossover of GA9899750.741001001009899830.82100100100
    (0.38–1.00)       (0.58–1.00)    
Equal flow in both arches9899830.821001001009999940.94100100100
    (0.58–1.00)       (0.81–1.00)    
Table 2. Interobserver agreement for evaluation of fetal cardiac structures, using two-dimensional (2D) and 4D ultrasound (US)
Cardiac parameter2D-US4D-US
Po (%)Ppos (%)Pneg (%)k (95%CI)Po (%)Ppos (%)Pneg (%)k (95%CI)
  1. 3VT, three vessels and trachea; Ao, aortic; AV, atrioventricular; Confl. arches, confluence of arterial arches on left of spine; GA, great arteries; IVS, intact ventricular septum; k, kappa; LA, left atrium; LVOT, left ventricular outflow tract; P0, overall proportion of agreement (number of readings (both positive and negative) on which both observers agreed, divided by total number of all readings performed (where total number = 100)); Ppos, positive agreement (number of positive findings on which both observers agreed divided by total number of positive readings recorded); Pneg, negative agreement (number of negative findings on which both observers agreed divided by total number of negative readings recorded); PV, pulmonary vein; SAC, septoaortic continuity.

Gray-scale        
Abdominal situs100100100100100100
Four-chambered heart100100100100100100
Intact crux9898970.96100100100
    (0.90 –1.00)    
At least one PV entering LA9898980.96100100100
    (0.90–1.00)    
LVOT and SAC9999970.96100100100
    (0.88–1.00)    
Confl. arches9999990.98100100100
    (0.94–1.00)    
Color Doppler        
Equal AV flow9798860.849899900.89
    (0.66–1.00)   (0.74–1.00)
IVS9798770.759798860.84
    (0.48–1.00)   (0.66–1.00)
Ao flow9899860.859899900.89
    (0.64–1.00)   (0.74–1.00)
Crossover of GA100100100100100100
Equal flow in both arches9999910.909698800.78
    (0.72–1.00)   (0.57–0.99)

Overall, the interobserver agreement was > 96% for all features evaluated, for both methods. For all 11 cardiac features, for both 2D- and 4D-US, there was good, excellent or perfect interobserver agreement (kappa > 0.6, Table 2). There was perfect interobserver agreement on 2D-US for two gray-scale parameters (abdominal situs and four-chambered heart) and for one color Doppler parameter (crossing of the great arteries), and on 4D-US for all gray-scale features and one color Doppler feature (crossing of the great arteries). There was excellent interobserver agreement (kappa > 0.8) on 2D-US for four gray-scale parameters (intact crux, at least one pulmonary vein entering the left atrium, LVOT and SAC, confluence of the arches) and three color Doppler parameters (equal atrioventricular flow, aortic flow and equal flow in both arches), and on 4D-US for three color Doppler parameters (equal atrioventricular flow, intact ventricular septum and aortic flow). There was good interobserver agreement (kappa > 0.6) for the one remaining color Doppler feature for each method: intact ventricular septum on 2D-US and equal flow in both arches on 4D-US.

Comparison of 2D-US and 4D-US for the visualization of normal fetal cardiac features is presented in Table 3. Fetal cardiac features seemed to be better assessed using 2D-US, but for Observer 1 the difference between techniques was statistically significant for only the LVOT and SAC and for Observer 2 there were no significant differences between 2D-US and 4D-US regarding visualization of fetal cardiac features.

Table 3. Visualization of cardiac anatomical structures by two-dimensional (2D) vs 4D ultrasound, by Observer 1 and Observer 2
Cardiac parameter Observer 1Observer 2
4D+4D–Pa4D+4D–Pa
  1. a

    McNemar's test comparing visualization of fetal heart anatomical structures with 2D or 4D ultrasound, rated as satisfactory (2D+, 4D+) or unsatisfactory (2D–, 4D–). 3VT, three vessels and trachea; Ao, aortic; AV, atrioventricular; Confl. arches, confluence of arterial arches on left of spine; GA, great arteries; IVS, intact ventricular septum; k, kappa; LA, left atrium; LVOT, left ventricular outflow tract; NS, not significant (P > 0.05); PV, pulmonary vein; SAC, septoaortic continuity.

Gray-scale       
Abdominal situs       
 2D+952NS952NS
 2D–03 03 
Four-chambered heart       
 2D+943NS943NS
 2D–03 03 
Intact crux       
 2D+594NS574NS
 2D–037 237 
At least one PV entering LA       
 2D+550NS530NS
 2D–045 245 
LVOT and SAC       
 2D+8060.031796NS
 2D–014 114 
Confl. arches       
 2D+532NS522NS
 2D–045 145 
Color Doppler       
Equal AV flow       
 2D+910NS862NS
 2D–09 39 
        
IVS2D+914NS866NS
 2D–05 26 
Ao flow       
 2D+913NS875NS
 2D–06 26 
Crossover of GA       
 2D+952NS952NS
 2D–03 03 
Equal flow in both arches       
 2D+923NS877NS
 2D–05 15 

Comparing grayscale with color Doppler imaging (Table 4), for both observers, on both 2D-US and 4D-US, we found statistically significant differences between results for two pairs of features: (1) LVOT and SAC/aortic flow and (2) confluence of arterial arches on left of spine/equal flow in both arches, with visualization on color Doppler being significantly better than that on gray-scale imaging. For both observers, for the pair of cardiac features assessed in the four-chamber view (four-chambered heart/equal atrioventricular flow), gray-scale imaging seemed to provide better visualization, without reaching statistical significance on 4D-US and on 2D-US for Observer 1, and reaching statistical significance on 2D-US for Observer 2.

Table 4. Visualization of cardiac anatomical structures by Observers 1 and 2 on two-dimensional (2D) and 4D ultrasound, comparing gray-scale vs color Doppler imaging
Gray-scaleObserver 1Observer 2
Color DopplerPaColor DopplerPa
  1. a

    McNemar's test comparing visualization of fetal heart anatomical structures using gray-scale or color Doppler imaging, rated as satisfactory (+) or unsatisfactory (–). NS, not significant (P > 0.05). AV, atrioventricular; Confl. arches, confluence of arterial arches on left of spine; LVOT, left ventricular outflow tract; SAC, septoaortic continuity.

2D      
 Equal AV flow +Equal AV flow – Equal AV flow +Equal AV flow – 
Four-chambered heart +898NS86110.022
Four-chambered heart –21 21 
 Aortic flow +Aortic flow – Ao flow +Aortic flow – 
LVOT and SAC +8600.0088410.039
LVOT and SAC –86 87 
 Equal flow in both arches +Equal flow in both arches – Equal flow in both arches +Equal flow in both arches – 
Confl. arches +550< 0.0001531< 0.0001
Confl. arches –405 415 
4D      
 Equal AV flow +Equal AV flow – Equal AV flow +Equal AV flow – 
Four-chambered heart +886NS868NS
Four-chambered heart –33 33 
 Aortic flow +Aortic flow – Aortic flow +Aortic flow – 
LVOT and SAC +7820.0077640.049
LVOT and SAC –137 137 
 Equal flow in both arches +Equal flow in both arches – Equal flow in both arches +Equal flow in both arches – 
Confl. arches +521< 0.0001485< 0.0001
Confl. arches –407 407 

DISCUSSION

In this study we investigated the feasibility and reproducibility of a traditional method, 2D-US (standardized cine-loop), and that of the relatively new 4D-STIC method to confirm normality of the heart in the first trimester, by visualizing as many fetal cardiac parameters as possible among a set of 11 features important in the confirmation of normal heart anatomy. Only a few authors have focused on confirming or refuting normality in the first trimester in a low-risk population[5, 7, 16, 19, 25].

Our results indicate high intra- and interobserver agreement in the assessment of all cardiac anatomical structures at 12 to 13 + 6 weeks' gestation for both 2D-US and 4D-US. There was good or excellent interobserver agreement for all the evaluated structures and the overall intraobserver agreement was > 95% for each observer using both methods. In terms of visualization rate, we found that fetal cardiac features were apparently better assessed using 2D-US compared with 4D-US, but the difference was not statistically significant. We also found a high visualization rate for the searched parameters, especially key features such as equal atrioventricular flow, crossover of the great arteries and equal flow in both arches, while using color flow mapping with both 2D and 4D methods. For the five-chamber view plane and three vessels and trachea view plane, color Doppler added valuable information regarding the outflow tracts. In the four-chamber view plane, the four-chambered heart on gray-scale was slightly better visualized than was equality of atrioventricular flow on color mapping. However, the latter added important information in confirming septum and symmetry of the ventricles, given the poor visualization of intact crux cordis on gray-scale imaging.

A strength of this study was our use of a standardized, homogeneous protocol, similar to the second-trimester assessment, that was relatively easy to perform and allowed comprehensive early cardiac evaluation of a large series of fetuses. 2D-US examinations are usually regarded as being observer-dependent, but the use of cine-loops acquired in a standardized way may alleviate this issue.

A limitation of this study was that, according to the study design, all examined fetuses had normal hearts; accuracy for detection of CHD thus cannot be evaluated from these data, for either of the methods. To confirm our data regarding observer agreement, larger studies are required, with increased number of observers and/or heart scans; this would also allow analysis of the bias of individual observers. A greater number of ultrasound examinations performed by two observers only would certainly narrow the 95% CIs around estimates of kappa, while having more than two observers would complicate the statistical analysis. Additionally, this was a single-center study with experienced operators only. Analysis involving operators with different levels of skill and from different centers is needed to determine whether our results are generalizable.

The good or excellent intra- and interobserver agreement that we found in both 2D- and 4D-US is in line with other research[4, 18]. Our results are also in agreement with those of other studies in demonstrating that these methods are reliable tools for early reassurance of normal cardiac anatomy or early suspicion of abnormal cardiac anatomy[4, 15, 19, 25, 29]. The high degree of visualization on 2D color Doppler imaging of the crossing of the great arteries and of flow in both arches is consistent with previous studies[4-6, 13] and support the early accessibility of these key cardiac planes. We consider our protocol to be readily achievable, as all 11 features were visualized satisfactorily (complete assessment) in more than half of the cases and all five color Doppler parameters were demonstrated in most cases.

Few studies have compared 2D with 4D-US for first-trimester cardiac assessment; in agreement with our findings, Bennasar et al.[14] obtained similar information with both methods. To our knowledge, there have been no studies comparing gray-scale and color Doppler techniques for visualization of the normal heart in the first trimester. However, the introduction of color/power Doppler to fetal cardiac investigation in the first trimester had a major impact on CHD detection rates[2, 4, 6, 7, 12, 13].

Our results imply that offline cardiac evaluation and expert reassessment is possible not only using 4D volumes, but also by reviewing 2D digital films that have been obtained in a standardized way. Moreover, digital films could serve as teaching material. Future research should investigate whether different levels of operator experience may influence the reproducibility of these methods.

Over a decade ago, it was considered inadvisable to perform early fetal cardiac screening[30]; however, more recent studies have shown routine 2D early fetal heart scanning to be possible[4-6, 13], with good visualization rates of cardiac features. In our view, a first-trimester cardiac screening protocol will eventually be established, as is the case for the second trimester[9, 10], that will result in earlier referral of suspected cases to tertiary centers, although second-trimester echocardiography is still considered and should remain the gold standard examination for the diagnosis of CHD. Second-trimester screening for CHD is no longer confined to analysis of the four-chamber view[8], nor should screening in the first trimester be so confined, given the high rate of visualization of structures in other planes and the high intra- and interobserver agreement found in our study and in previous publications.

Our data show a statistically significant improvement in visualization of cardiac features when using color Doppler compared with gray-scale imaging. It seems logical to use it as part of the 2D-US examination, given that similar information was obtained with 2D- and 4D-US, and that 2D-US is traditionally more widely used, being a cheaper and simpler method compared with 4D-US. However, future bioethical studies should examine the implications of early fetal assessment using color Doppler, and care should be taken in extrapolating the results of this study to the general population, superior-quality equipment and training being crucial in confirming or refuting normality in the very early stages of pregnancy.

Supporting Information On The Internet

Videoclips S1–S3 may be found in the online version of this article.

Ancillary