The role of echocardiography and magnetic resonance imaging in the diagnosis of foetuses and neonates with an antenatal suspicion of coarctation of the aorta.

Abstract

Background: Diagnosing coarctation of the aorta (CoA) in early life is difficult, particularly in neonates in whom ductal patency can obscure anatomical narrowing, and anatomical landmarks alone may not reflect functional obstruction. This thesis had two aims: (1) to systematically review, appraise and externally validate existing foetal and neonatal anatomical echocardiography-based prediction models for CoA; and (2) to evaluate the incremental diagnostic value of neonatal feed-and-wrap cardiac magnetic resonance imaging (CMR) as a complementary modality that integrates functional and haemodynamic information.

Methods: Study 1 was a systematic review of foetal and early neonatal echocardiographybased CoA prediction models. Study 2 was a retrospective external validation study that validated selected foetal and neonatal echocardiography prediction models. Study 3 was a prospective study of neonates with antenatal suspicion of CoA who underwent non-sedated, non-contrast feed-and-wrap CMR to assess left ventricular (LV) function, myocardial deformation, and advanced haemodynamic parameters using 4D Flow CMR.

Results: Study 1: Twenty-one studies were identified: Internal validation was largely absent, none performed external validation, calibration was rarely assessed, clinical utility was seldom examined and overall risk of bias was high (PROBAST). Study 2: Four echocardiographic models assessed in a retrospective cohort of 190 foetuses and neonates. All four models showed poor calibration in their original form. After updating these models, the combined foetal and neonatal model showed the best overall performance with strong discrimination and higher net benefit across the full range of threshold probabilities. Study 3: Neonates were categorised into three groups based on postnatal diagnosis: Non-CoA, hypoplastic aortic arch, and confirmed CoA. CMR-derived LV volumetrics were highly reproducible. CMR feature-tracking–derived LV strain was significantly lower in CoA and hypoplastic-arch groups than in false-positive cases, with strong inter observer/inter-vendor agreement and excellent diagnostic accuracy (AUC 0.917–0.975). In the same study, 4D Flow CMR (including super-resolution) was feasible and clinically integrable. Pressure-based metrics and flow-disturbance markers discriminated CoA with high accuracy (AUCs ≥0.94) and were robust at native and super-resolution. Energy-loss and viscous-dissipation had moderate discrimination. Wall shear stress was resolution-sensitive.

Conclusion: Current foetal and neonatal echocardiography-based prediction models need more rigorous development, reporting, and external validation. Feed-and-wrap CMR in neonates is feasible and yields reproducible functional and haemodynamic measures that can complement echocardiography. Integrating these measures with anatomical parameters may improve future risk stratification tools.