A Multiparametric, Reliability‐Weighted Fetal Brain Biometry Portrait for the Prediction of Multi‐Domain Neurodevelopmental Outcomes Across 12 and 24 Months of Age

Prenatal brain structural MRI opens a pioneering window on the identification of brain-behavior relationships. In this work, we aim to establish the reliability of fetal brain structural biometric parameters for delineating a "fetal biometry portrait." Fetal brain biometric parameters were assessed in a sample of 202 fetuses with typical structural brain development scanned between 21.3 and 36.7 gestational weeks and encompassed (i) local and global cortical, (ii) posterior fossa, and (iii) deep gray matter nuclei development. Prenatal brain MRI scoring was performed by four independent raters, and reliability was assessed. Neurodevelopment was assessed at 12 and 24 months, using the Bayley-III scales (BSID-III). A convex relaxed clustered multi-task learning (CMTL) model, was used (i) to test the fetal brain biometry "reliability-weighted" model performance, (ii) to identify a shared clustered structure among the different BSID-III domains, (iii) to mark clusters incorporating multiparametric features ensembles of fetal structural maturation, and (iv) to characterize biometrical features embedded in specific clusters or shared between clusters for the prediction of neurodevelopmental outcomes in all BSID-III domains across timepoints. Results showed that the CMTL 'reliability-weighted' model achieved a good performance. Four clusters were identified, grouping patterns of BSID-III domains. Multiparametric ensembles were identified with cluster-specific and clusters-shared top-ranked features.