Optimisation of fetal strain by CMR and its use in suspected coarctation of the aorta

Background: Coarctation of the aorta (CoA) is associated with high morbidity and mortality, if not detected in time. Further, prenatal diagnosis has low detection rates and high false positive rates. In newborns with suspected CoA, reduced left ventricular strain has been shown in true positive cases compared to false positive cases. It is unclear whether this difference is present already prenatally. Fetal CMR has assessed strain in cases of suspected congenital heart defects, and may provide diagnostic information in cases where echocardiography is inconclusive, including CoA. However, the use of strain by fetal CMR has yet to be improved and validated.

This study therefore aimed to 1) optimise left ventricular fetal strain assessment; and 2) explore the potential of fetal strain by CMR to differentiate between true and false positive cases of CoA.

Methods: Fetal strain was assessed in 32 fetuses during the third trimester and compared with postnatal outcome (9 true positive CoA, 32–39 gestational weeks; 16 false positive; 32–40 gestational weeks; 8 controls, 31–39 gestational weeks). Cine images gated to the cardiac cycle by Doppler ultrasound (Smartsync, Northh, Hamburg, Germany) were acquired at 1.5 T (Siemens Aera, Siemens Healthineers, Forchheim, Germany). For left ventricular strain feature tracking was applied and for right ventricular strain endocardial shortening was assessed (Segment v4.1.0.0 R15011, Medviso, Lund, Sweden). Optimisation of left ventricular strain was performed both separately in long-axis images and for global longitudinal strain by systematically adjusting parameters for contour stiffness and tracking distance between timeframes. To determine the optimal parameter settings, both the visual accuracy of the automatic tracking and change in strain values were assessed. Right ventricular strain optimisation was inferred from these values, and optimised parameter sets were used for both left and right ventricular strain assessment in the fetuses with true CoA, false positive CoA, and healthy controls.

Results: Figure 1 shows impact of parameter optimisation on strain measurements. Figure 2 shows difference in 4-chamber strain before and after parameter optimisation (bias 2.3 %-age units; 95 % limits of agreement 0.6–6.9 %-age units; p< 0.001). Figure 3 shows left and right ventricular strain for fetuses with true positive CoA, false positive CoA, and healthy controls.

Conclusion: Optimisation of left ventricular strain assessment impacts strain values. It is likely that right ventricular strain optimisation shows a similar trend. Therefore, further optimisation by large-scale computer phantoms and comparison to other methods are needed for validation of strain values by fetal CMR. Nevertheless, a tendency in decreasing left ventricular strain may potentially, together with other imaging biomarkers, help separate true CoA and false positive CoA already prenatally.

Figure 1. Impact of parameter optimisation on left ventricular longitudinal strain values. For both contour stiffness and tracking distance, parameter values closer to zero yielded a more lenient tracking while higher values yielded a stiffer tracking. The optimal values, including by visual assessment of tracking in the cine images for the different parameter settings, were 0.3 for contour stiffness and 3 mm for tracking distance. Triangles denote 4-chamber strain, boxes denote 3-chamber strain, and filled circles denote global longitudinal strain.
Figure 2. Bland-Altman plot showing the agreement between left ventricular strain values before and after optimisation. Before optimisation, left ventricular 4-chamber strain was -12 % [-10 to -15], which changed to -16 % [-12 to -18] after optimisation (p < 0.001), with bias 2.3 %-age units and 95 % limits of agreement 0.6–6.9 %-age units.
Figure 3. Longitudinal 4-chamber strain for the left and right ventricles in fetuses developing CoA, fetuses that were false positive, and in healthy controls. Although a visual tendency, trend analysis show that there is no trend of less negative left ventricular strain values in fetuses that developed CoA after birth compared to false-positive cases and controls (p=0.07).