(70) Unsupervised Machine Learning of Fetal Myocardial Performance Index Trajectories Identifies FGR and SGA Patterns

Objective:
To use unsupervised machine learning approaches on longitudinal trajectories of fetal left ventricular myocardial performance index (MPI) and its association with the risk of subsequent fetal growth restriction (FGR) and small for gestational age (SGA) birthweight.

Study Design:
This was a prospective cohort study which enrolled patients at ≤ 13+6 weeks’ gestation without fetal, umbilical cord or placental abnormalities. Left ventricular MPI was measured once per trimester using tissue Doppler at 12-13+6, 20-21+6- and 32-33+6-weeks’ gestation. FGR was defined as estimated fetal weight < 10th percentile for gestational age. SGA was defined as birthweight < 10th percentile for gestational age. Gaussian Mixture Models clustering was applied to identify distinct functional trajectories for left ventricle MPI. Rates of FGR and SGA neonates were compared across clusters using chi-squared test.

Results:
Among 306 pregnancies, three distinct MPI trajectories were identified (Figure 1). Cluster 1 consisted of fetuses with low first-trimester MPI followed by a delayed rise in the third trimester. Cluster 2 included fetuses with persistently elevated MPI across all three time points. Cluster 3 represented fetuses with low first-trimester MPI that demonstrated early recovery by the second trimester. FGR was more common in Cluster 1 (11.6%) compared to Clusters 2 (2.8%) and 3 (5.4%) (p=0.031). SGA was also most frequent in Cluster 1 (7.2% vs. 0.7% and 0.0%; p=0.001).

Conclusion:
FGR is characterized by early pregnancy low MPI that does not recover by the second trimester. These findings suggest that fetal cardiac function adapts to an early pregnancy event, likely utero-placental insufficiency, and that failure of this adaptation is associated with FGR.