Role of four-dimensional ultrasound (spatiotemporal image correlation and Sonography-based Automated Volume Count) in prenatal assessment of atrial morphology in cardiosplenic syndromes

Authors


Abstract

Objective

To assess the diagnostic role of four-dimensional ultrasound using spatiotemporal image correlation and Sonography-based Automated Volume Count (STIC-SonoAVC) in the identification of the morphology of the atrial appendages in cases with cardiosplenic syndrome.

Methods

This was a retrospective investigation of 22 fetuses with cardiosplenic syndromes seen at our institution over a 5-year period from January 2004. As control groups, 10 normal fetuses, five cases with a non-isomeric atrioventricular septal defect and five cases with other congenital heart diseases were also analyzed. For all fetuses, one or more cardiac volume datasets were available for offline analysis. Two-dimensional and four-dimensional echocardiography was carried out in all cases at the time of diagnosis using high quality three-dimensional equipment. Dedicated software was used to assess chamber morphology using the SonoAVC technique, which allows the creation of casts of hollow structures. Two different operators used the software. The first performed all steps up to positioning of the region of interest box. The second operator, who was blinded to clinical information, then rendered the cardiac chambers using the SonoAVC technique. This operator then used the rendered image to subjectively assess atrial morphology.

Results

Suitable rendered images of the cardiac chambers could be produced in 40/42 fetuses. In two cases of left atrial isomerism, advanced (34 weeks) and early (13 weeks) gestational age made it impossible to obtain adequate rendered images. In the remaining 40 cases (13 cases of left atrial isomerism, seven cases of right atrial isomerism, five cases of non-isomeric atrioventricular septal defect, five cases of other congenital heart diseases and 10 normal fetuses), atrial morphology was correctly identified by evaluation of the rendered images.

Conclusion

Four-dimensional ultrasound with SonoAVC rendering allows correct identification of the morphology of atrial appendages in all cases of cardiosplenic syndromes in which an adequate cardiac volume dataset can be obtained for analysis. Copyright © 2011 ISUOG. Published by John Wiley & Sons, Ltd.

INTRODUCTION

Cardiosplenic syndromes are complex multiorgan conditions defined by the arrangement of the thoracic and abdominal viscera across the left–right axis1, 2. In left isomerism, there are paired left-sided viscera while the right-sided viscera may be absent, and in right isomerism, there are paired right-sided viscera with possible absence of left-sided viscera3, 4. In the former condition, the classic features are bilateral morphological left atrial appendages (left atrial isomerism), multiple cardiac anomalies, bilateral morphological left (bilobed) lungs with hypoarterial bronchi, polysplenia, intestinal malrotation and interruption of the inferior vena cava with azygos continuation4–12. Conversely, in right isomerism typical anomalies are bilateral morphological right atrial appendages (right atrial isomerism), multiple cardiac anomalies, bilateral morphological right (trilobed) lungs with epiarterial bronchi, asplenia and a malpositioned inferior vena cava, which may be anterior or juxtaposed to the aorta4–13.

Recognition of heterotaxy in the fetus is often challenging because prenatal evaluation of lung lobulation, the bronchial branching pattern, spleen status and morphology of the atrial appendages is difficult.

It has been shown that the best indicator of isomerism in children with congenital heart disease is the shape of the pectinate muscle14. In the fetus, autopsy studies have demonstrated that the best indicator of the presence of isomerism is identification of the morphology of the atrial appendages8. Evaluation of atrial morphology (rather than atrial appendage morphology) has also been described, but its use is also questionable and has not been reproduced by other groups13. It would therefore be beneficial to have a technique that could be used to reliably and easily identify the morphology of atrial appendages in the fetus.

Spatiotemporal image correlation (STIC) is a technique that allows the acquisition of cardiac volumes15. Information extracted from these datasets can be displayed and analyzed in multiplanar or rendering modes, including Sonography-based Automated Volume Count (SonoAVC)16, which has also been employed in fetal cardiology to display casts of the heart and great vessels17.

We have used this technique to study the morphology of the atrial appendages in heterotaxy. The aim of this retrospective study was to define the diagnostic accuracy of STIC-SonoAVC in the identification of atrial morphology in cardiosplenic syndromes in the fetus.

METHODS

This was a retrospective investigation that included all cases of confirmed cardiosplenic syndromes managed at our unit in the 5-year period from January 2004 in which one or more cardiac volume datasets had been acquired and stored. In this period, we had evaluated 25 fetuses with cardiosplenic syndromes, and cardiac volume datasets were available for 22. For comparison, atrial morphology was assessed also in a series of 10 fetuses with normal hearts, five fetuses with other cardiac abnormalities and five fetuses with non-isomeric atrioventricular septal defect (AVSD). The five cases with other heart diseases included one case each of ventricular septal defect, aortic coarctation, corrected transposition, tetralogy of Fallot and transposition of the great arteries.

Two-dimensional (2D) and four-dimensional (4D) echocardiography was carried out in these 42 fetuses using high-quality three-dimensional (3D) equipment (Voluson 730 Expert and E8/E9; GE Medical Systems, Zipf, Austria). One or more cardiac volume datasets were acquired using a transabdominal volumetric 4–8-MHz transducer and stored in an archive. For this study, dedicated software was used (4D View version 10.0) by two different operators, with the latter blinded to clinical information. For each fetus, the first operator performed the following steps: (i) removal of patient details from the screen; (ii) selection of the best gray-scale volume for assessing anatomy in the four-chamber view; (iii) selection of the best frame of the systolic phase for analysis, using the cineloop function; (iv) adjustment of the image brightness and contrast to enhance automated border recognition between cavities and the myocardium; and (v) positioning of the region of interest (ROI) box, which was set to completely include the atria and the ventricles. Once the ROI was fixed, the second operator performed the following steps: (i) creation of atrial and ventricular casts by clicking into the various cardiac chambers; (ii) use of multiplanar correlation analysis to correct for possible ‘bleeding’ of the color outside the cardiac structures as a result of shadowing from ribs or limbs; and (iii) arbitrary separation of atria from ventricles using the dedicated function if there was bleeding of color across the atrioventricular valves. The rendered image was then used by the second operator for the subjective assessment of the morphology of the atrial appendages. The time taken to reach the final stage from the initial opening of the volume (i.e. including the processes performed by both operators) was also recorded.

Final confirmation of the diagnosis was available for all 42 fetuses, by means of necropsy reports (in cases of termination of pregnancy) or the findings on postnatal catheterization, surgery or ultrasound (in cases of livebirths).

RESULTS

The median gestational age at diagnosis was 22 (range, 13–34) weeks. Overall, there were 15 cases of left atrial isomerism and seven cases of right atrial isomerism. There was no difference in mean gestational age among the four groups (isomerism, normal hearts, non-isomeric AVSD or other heart defects). The various cardiac anomalies detected in the 22 cases of isomerism are summarized in Table 1. Suitable rendered images of the cardiac chambers could be produced in all but two cases (including all cases with a normal heart, 5/5 cases with non-isomeric AVSD, 5/5 cases with other heart defects and 20/22 cases with isomerism). In one case diagnosed with left atrial isomerism at 13 weeks (single ventricle + malposition of great vessels + pulmonary outflow obstruction + complete heart block and enlarged nuchal translucency), the quality of the volume was suboptimal as a result of heart block and early gestational age, and we were not able to use the SonoAVC function to produce casts of the cardiac chambers. For the second case, it was again impossible to obtain adequate rendered images as a result of significant shadowing from the ribs, which was related to the advanced gestational age (34 weeks). In the remaining 20 cases with cardiosplenic syndromes, atrial morphology was correctly identified by evaluation of the rendered images. In Figure 1, examples of the morphology of the atrial appendages, as seen on STIC-SonoAVC in four cases of left atrial isomerism and in three cases of right atrial isomerism, are shown, together with a normal situs solitus reference case. In Figure 2, another normal case (a) is shown together with cases of corrected transposition (b) and aortic coarctation (c) and three cases of non-isomeric AVSD (d, with unbalanced ventricles; e and f, with balanced ventricles).

Figure 1.

Images generated using spatiotemporal image correlation and Sonography-based Automated Volume Count, showing morphology of the atrial appendages in normal situs solitus (a), left isomerism (b–e) and right isomerism (f–h). Note the digit-like shape of the appendage which identifies the atrium as left-sided (arrows) in comparison with the wide and broad-based shape of the right-sided appendage (arrowheads). Concurrent lesions were: unbalanced atrioventricular septal defect (b, c, f, h) and single ventricle (d, e, g). All cases were between 20 and 25 weeks of gestation. Numbers and colors are assigned automatically by the software to individual hollow structures.

Figure 2.

Images generated using spatiotemporal image correlation and Sonography-based Automated Volume Count showing that the differential morphology of atrial appendages (the arrow indicates the digit-like left-sided appendage) can be clearly recognized in different types of congenital heart disease. (a) Normal situs solitus. (b) Corrected transposition of the great arteries; note the atrioventricular discordance, with the morphological right ventricle (mRV) on the left side, under the left-sided atrium with the digit-like appendage (arrow). The arrowhead identifies the moderator band. (c) Aortic coarctation; note the ventricular disproportion, with the right ventricle larger than the left. (d) Non-isomeric atrioventricular septal defect (AVSD) with unbalanced ventricles. (e, f) Non-isomeric AVSD with balanced ventricles. CA, common atrium; LV, left ventricle; RA, right atrium.

Table 1. Cardiac malformations detected in the 22 cases of heterotaxy analyzed in this study
CaseGestational age (weeks)Type of isomerismCardiac lesions
  • *

    Cases not assessed using Sonography-based Automated Volume Count (Sono-AVC).

  • Ao, abdominal descending aorta

  • AVSD, atrioventricular septal defect, unbalanced

  • CHB, complete heart block

  • DORV, double-outlet right ventricle

  • GA, great arteries

  • IVC, inferior vena cava

  • Pa, pulmonary artery

  • PAPVR, partial abnormal pulmonary venous return

  • SV, single ventricle

  • TAPVR, total abnormal pulmonary venous return.

1*13LeftIVC interruption + azygos continuation + SV + GA malposition + Pa stenosis + CHB
217LeftIVC interruption + azygos continuation + SV + GA malposition + Pa stenosis + CHB
322LeftIVC interruption + azygos continuation + SV + GA malposition + Pa stenosis + CHB
425LeftIVC interruption + azygos continuation + SV + GA malposition + Pa stenosis + CHB
521LeftIVC interruption + azygos continuation + AVSD + DORV + Pa stenosis + CHB
628LeftIVC interruption + azygos continuation + AVSD + DORV + Pa stenosis
721LeftIVC interruption + azygos continuation + AVSD + DORV + Pa atresia + CHB
820LeftIVC interruption + azygos continuation + AVSD + DORV + Pa atresia
915LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
1023LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
1121LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
12*34LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
1327LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
1418LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
1516LeftIVC interruption + azygos continuation + AVSD + GA malposition + Pa stenosis
1621RightJuxtaposition IVC Ao + TAPVR + SV + GA malposition + Pa stenosis
1721RightJuxtaposition IVC Ao + PAPVR + SV + GA malposition + Pa stenosis
1825RightJuxtaposition IVC Ao + PAPVR + SV + GA malposition + Pa stenosis
1923RightJuxtaposition IVC Ao + PAPVR + AVSD + GA malposition + Pa atresia
2026RightJuxtaposition IVC Ao + AVSD + GA malposition + Pa atresia
2117RightJuxtaposition IVC Ao + AVSD + DORV + Pa stenosis
2221RightJuxtaposition IVC Ao + AVSD + DORV+ Pa stenosis

The accuracy of the procedure was high; in the 40/42 fetuses in which all steps could be carried out, the diagnosis of the morphology of the atrial appendages exactly matched the final diagnosis, as reported on necropsy or postnatally.

The median time required to obtain the rendered images (i.e. the total time for the two operators) was 5 (range, 3–15) min. A median of three (range, 2–10) regions of color bleeding required correction. Color bleeding occurred more frequently for volumes obtained before 17 and after 27 weeks of gestation.

DISCUSSION

Recognition of heterotaxy in individuals with congenital heart disease is a challenging task, even at necropsy14, as the characteristics of the atrial appendages are not always easily described by the pathologist14. In postnatal life, the shape of the pectinate muscle has been advocated to be a better identifier of isomerism than is the morphology of the atrial appendages14, but this anatomic structure is very difficult to identify and assess in the small fetal heart, even with 3D/4D ultrasound technology. Hence, in the fetus, the morphology of the atrial appendages remains the most reliable indicator of the presence and the type of isomerism8. However, it is difficult, and often impossible, to recognize this feature on 2D imaging. As the two atria differ not only in the morphology of the appendages but also in that of the atria themselves, there have been attempts to use this type of evaluation to differentiate the left-sided from the right-sided atrium13, but these preliminary results have not been validated by other researchers. The problem of identification of the morphology of atrial appendages in fetuses with cardiosplenic syndromes has therefore remained open until now.

STIC-SonoAVC is a technique based on edge recognition which fills in hollow structures with color, creating casts of the structures. We decided to investigate if this tool could be used to characterize the morphology of the atrial appendages in fetuses with cardiosplenic syndromes. However, considering that evaluation of the atrial appendages is not always straightforward in hearts with situs solitus14, we decided to include in the study 10 fetuses with normal hearts and five with non-isomeric AVSD to further explore the accuracy of the technique. As evident from Figures 1 and 2, the technique is effective, with both the digit-like narrow-based left-sided appendages and the wide broad-based right-sided appendages clearly demonstrated in color both in fetuses with situs solitus (Figure 1a and Figure 2a) and in those with cardiosplenic syndromes (Figure 1b–h).

It needs to be stated that there are some technical limitations to the method described, the most frequent being the bleeding of color outside the cardiac chambers. However, it was possible to resolve this problem using multiplanar correlation to cut out the non-cardiac signal in 20/22 cases with cardiosplenic syndromes (Figure 3). With a few clicks, using the multiplanar correlation and the ability of the software to cut and remove colored areas from the three orthogonal planes, the artifacts could be removed and the actual shapes of the atrial appendages revealed. The need to edit the regions showing bleeding of color was the main factor accounting for the time spent using the software. If only cases with two areas of bleeding are considered, the mean time needed to complete the rendering from the opening of the volume was only 3 min.

Figure 3.

When the automated edge recognition of the Sonography-based Automated Volume Count software fails as a result of shadowing from ribs or from failure to detect the fluid–solid border (as in this case), the color bleeds outside the cardiac chambers into the myocardium or thorax. Using the multiplanar correlation, this artifact can be detected and removed. In (a) the bleeding, in yellow, can be recognized both on the rendered image (lower right panel, arrows) and on two of the three orthogonal planes (arrowheads). The problem can be identified and resolved by manual editing using the software (b). Numbers and colors are assigned automatically by the software to individual hollow structures.

Another technical issue to consider is the fact that in some cases the left-sided small appendages did not fill with color on the first attempt; if the operator was not careful in evaluating this on the three planes, the incompletely rendered image could be misleading, showing false right-sided appendages (Figure 4). However, in no case did this actually lead to misjudgement of the type of isomerism as the operator was able to identify the problem and add the appendage to the main atrial volume with a second click.

Figure 4.

In some cases the left-sided small appendages (arrow in A-plane of (a)) did not fill with color on the first attempt using the Sonography-based Automated Volume Count software. If the operator is not careful in evaluating this on the three planes, the incompletely rendered image could be misleading, showing false right-sided appendages (a; arrowheads). However, in no case did this actually lead to misjudgement of the type of isomerism because the operator was able to identify the problem and add the appendage to the main atrial volume with a second click, showing its left-sided morphology (b; arrows). Note that the position in space of the rendered cast of (b) is reversed in comparison with that in (a). Numbers and colors are assigned automatically by the software to individual hollow structures.

A final technical limitation is the fact that the software does not allow the operator to assign the colors from a palette, but assigns them by default following a volume scale (largest volume in red, second largest in blue, etc.). This is not ideal because the single chambers cannot be tagged using a specific color code. However, this inconvenience could be resolved by the manufacturer of the software.

A possible criticism of this study is that 3D technology is not needed to describe atrial appendages, considering that we used, as reference images, the three orthogonal 2D planes to correct for bleeding of color outside the atrial cavities. This is only apparently true for two reasons: (i) we used the multiplanar imaging mode only to display the borders of the atrial cavities, but it would not be possible to derive with certainty the shape of the whole structure using this approach; and (ii) multiplanar imaging is a 3D technology anyway.

Another possibility may be to use SonoAVC to identify the pectinate muscle, which was demonstrated in postnatal life to be a better indicator of isomerism than the atrial appendage8. However, SonoAVC is used to visualize casts of hollow structures (atria and ventricles); the pectinate muscle would therefore be shown as an indentation in the virtual atrial cast. To assess such an indentation in a solid cast would be very difficult or impossible, particularly given the small dimensions of the fetal heart.

There are also some minor conceptual limitations of this study. The first criticism that can be raised is that experienced fetal cardiologists may be able to define the morphology of atrial appendages using 2D imaging alone. However, there are no studies in the literature describing this and other expert researchers have concluded that this method of diagnosis would be unreliable13. Another criticism may be that the results were good because the volumes obtained were of good quality and that the same results need to be reproducible by other investigators. Although this may be the case, fetal cardiology units, by definition, should have the expertise required to obtain volumes of similar quality.

In conclusion, we have shown that STIC-SonoAVC is able to depict, with high accuracy, the morphology of atrial appendages in the normal fetal heart, in cases of non-isomeric AVSD and in fetuses with cardiosplenic syndromes, especially in the second trimester. In our opinion, this finding is of the utmost importance because it allows the presence and the type of isomerism to be defined in those cases in which doubts persist after 2D echocardiography. We hope that in the near future our findings will be reproduced by other investigators working in fetal cardiology.

Ancillary