Right ventricular (RV) function and its quantitative assessment, particularly volumetric analysis, are well-established markers associated with prognosis and the need for intervention in both congenital and acquired heart disease. Cardiac magnetic resonance (CMR) remains the reference standard for this evaluation. However, its use in children has important limitations related to availability, contraindications, and the need for sedation. Three-dimensional transthoracic echocardiography (3DTTE) has shown promising results in this setting, particularly in adults, in whom very strong correlations with CMR have been reported for RV volumes and right ventricular ejection fraction (RVEF). In children, however, evidence remains limited. Some studies include healthy children without heart disease, whereas others lack systematic comparison with CMR in all patients, particularly those evaluating automated 3DTTE software. We therefore considered it important not only to assess this technique in children, but also to explore its potential as an advanced echocardiographic tool in patients whose ventricular geometry is often complex.

To address these questions, we conducted a retrospective study involving a cohort of 92 children and adolescents with biventricular congenital or acquired heart disease who had undergone CMR and automated 3DTTE within a maximum interval of six months. Automated software was used to obtain the main parameters of RV size and function: indexed right ventricular end-diastolic volume (RVEDVi), indexed right ventricular end-systolic volume (RVESVi), RVEF, and linear measurements including TAPSE and basal RV diameter. These measurements were compared with those obtained by CMR, while TAPSE and basal RV diameter were also compared with measurements obtained by a pediatric cardiologist using two-dimensional transthoracic echocardiography (2DTTE). Statistical analysis included Pearson correlation coefficients to assess the relationship between techniques, Bland-Altman analysis to evaluate bias and limits of agreement, and kappa coefficients to determine agreement in ventricular function classification. In practical terms, Pearson correlation allowed us to determine whether both techniques behaved similarly and identified the same patients as having larger or smaller ventricles, whereas Bland-Altman analysis quantified how far automated 3DTTE measurements deviated from those obtained by CMR. The kappa coefficient was also used to assess whether both techniques agreed in classifying patients according to the presence or absence of RV dysfunction (RVEF <45%). In addition, following a suggestion from one of the reviewers of our manuscript, a second-stage analysis assessed the correlation between both techniques separately in patients with greater RV dilatation (≥100 mL/m²) and those with less dilatation, comparing the strength of the correlations between the two groups. Interobserver reproducibility was assessed in a random sample of 20 studies using intraclass correlation coefficients (ICC). Finally, we evaluated the clinical feasibility of the technique, including the need for manual corrections despite the automated nature of the software, the proportion of studies suitable for analysis, and the time required to complete automated quantification.

Of the 92 patients initially identified, 81 could be adequately analyzed. Automated 3DTTE showed high feasibility, with a mean analysis time of only approximately four minutes. From a practical standpoint, however, one of the most relevant findings was that the anatomical landmarks automatically identified by the software had to be verified by the operator in 94% of studies and corrected whenever necessary. Compared with CMR, automated 3DTTE showed a strong correlation for ventricular volumes (r = 0.81 for RVEDVi and r = 0.82 for RVESVi; 𝑝 < .001) and a moderate correlation for RVEF (r = 0.70; 𝑝 < .001). Although echocardiography tended to modestly underestimate volumes compared with CMR, with a mean bias of 15.1 mL/m² for RVEDVi and 7.8 mL/m² for RVESVi, no significant difference was observed for RVEF (𝑝 = .70). From a clinical perspective, the technique demonstrated excellent ability to identify patients with RV dysfunction. Using an RVEF cutoff of <45%, both techniques classified the vast majority of patients concordantly (κ = 0.82; 𝑝 < .001), with a negative predictive value of 100%.

The results remained consistent both in patients with more dilated RVs and in those with lesser degrees of dilatation. Fisher testing showed no significant differences in the strength of the correlations between the two subgroups. Interobserver reproducibility was excellent, with intraclass correlation coefficients of 0.92 for RVEDVi, 0.96 for RVESVi, and 0.93 for RVEF (𝑝 < .001). Automated 3DTTE measurements also showed very good correlation with those obtained by the pediatric cardiologist using 2DTTE for basal RV diameter (r = 0.98; 𝑝 < .001) and TAPSE (r = 0.82; 𝑝 < .001).

Taken together, these findings led us to conclude that automated 3DTTE provides good correlation with CMR for the assessment of RV volumes and RVEF, together with excellent reproducibility and clinical applicability for routine follow-up, including in pediatric patients.

COMMENTARY:

This study was conceived around two main objectives. The first was to assess the feasibility and accuracy of this noninvasive technique for RV evaluation in children with sometimes complex ventricular geometries using automated 3D software. This objective was achieved, as shown by the high correlation with the reference technique and by the fact that the analysis could be successfully completed in most patients. As expected, the correlations were lower than those reported in other studies using the same tool in adults and in structurally normal hearts. This is entirely reasonable, because endocardial border tracking is easier in such patients and most advanced echocardiographic tools incorporating artificial intelligence have historically been developed and trained predominantly in adults with more conventional ventricular anatomy. One of the strengths of our study was therefore the inclusion of one of the larger pediatric series in which every child underwent both automated 3DTTE and CMR.

Does this first objective have limitations? Certainly. The analysis does not overcome the requirement for adequate acquisition of the entire RV on 3DTTE, nor does it eliminate the need for some familiarity with adjustment of the anatomical landmarks used by the software. Automated analysis of abnormal ventricular geometries using artificial intelligence is clearly feasible, but expert supervision remains necessary to ensure that the software has correctly identified the relevant anatomical landmarks. A further limitation is that these findings should ideally be reproduced in an external cohort of children with heart disease to confirm their generalizability and reproducibility.

Our second objective was, I must admit, partly scientific and partly romantic: to challenge the repeatedly stated assumption that advanced echocardiographic techniques (and particularly those incorporating artificial intelligence) cannot be applied to pediatric cardiology. It is worth remembering that supervised software analysis showed a high correlation with CMR while providing bedside assessment in a noninvasive, accessible, rapid, and accurate manner. Following this study, we were able to introduce this tool into our echocardiography laboratory, although considerable work remains before it can be implemented extensively across other pediatric centers.

Finally, one important limitation—and certainly one of the lessons we learned—was highlighted by one of the journal reviewers. Dear readers, demanding reviewers improve our studies, even though the way they make their point may occasionally surprise us. Translated into gentler terms, the reviewer’s message was essentially this: stop validating quantitative 3D assessment of RV volumes and function solely against CMR and start defining pediatric-specific volumetric thresholds for this technique that are associated with prognosis or the need for intervention. And indeed, this study does not achieve that objective, which remains an important limitation. Mr Reviewer: you were right. Challenge accepted for the next study.

REFERENCE:

Deiros-Bronte L, Zabaleta MA, Buitrago NM, Uceda A, Diez-Sebastián J, Bret M, et al. Automated Three-Dimensional Transthoracic Echocardiography for Right Ventricular Assessment in Children with Congenital and Acquired Heart Disease: Validation against Cardiac Magnetic Resonance. J Am Soc Echocardiogr. 2026 Jul;39(7):692-694. doi: 10.1016/j.echo.2026.03.009.