Deep learning assessment of fetal brain maturation on 3D ultrasound volumes in early‐onset fetal growth restriction

ABSTRACT Objectives To quantify fetal brain maturation in fetuses with early‐onset fetal growth restriction (FGR) by estimating gestational age (GA) based on the appearance of gyrification and individual brain structures from three‐dimensional (3D) ultrasound volumes using a deep‐learning model trained on optimally developing subjects. The association of altered fetal brain maturation, as a potential marker of cumulative intrauterine stress, with an increased risk of neonatal complications was also explored. Methods This was a prospective, observational, single‐center cohort study of singleton pregnancies with early‐onset FGR conducted at the University Medical Center Utrecht between June 2022 and June 2024. Early‐onset FGR was defined as an estimated fetal weight (EFW) and/or abdominal circumference below the 10 th percentile before 32 weeks' gestation, and brain sparing was defined as an umbilical artery pulsatility index (PI) above the 95 th percentile in combination with a middle cerebral artery PI below the 5 th percentile or a cerebroplacental ratio (CPR) < 1. Fetal brain maturation was determined using a deep‐learning model, which was trained on data from an optimally developing cohort at GAs of between 18 + 0 and 28 + 6 weeks, collected by the INTERGROWTH‐21 st Consortium. Therefore, only fetuses with a GA of < 29.0 weeks at the time of the scan were included in the analysis. Estimation of brain maturation was based on the size and shape of the Sylvian fissure (SF), parieto‐occipital fissure (POF) and calcarine sulcus (CLC), cerebellum (CB) and the combination of all fissures using the whole‐brain ultrasound scan. The mean difference between estimated GA and actual GA (ΔGA) in days was calculated. In a subgroup analysis, ΔGA for brain‐sparing FGR was compared with ΔGA for non‐brain‐sparing FGR. Regression analysis was performed to examine the relationship between brain maturation data and potential covariates. Results The study included 43 growth‐restricted fetuses with high‐quality 3D ultrasound scans (13 of which had brain‐sparing FGR) at a median GA of 27.1 (interquartile range (IQR), 26.1–27.7) weeks. The estimated GA was significantly lower than the actual GA in early‐onset FGR, with ΔGA of –4.6 (IQR, –9.8 to –1.0) days based on the whole scan, ΔGA of –4.9 (IQR, –9.0 to –1.2) days based on the SF, ΔGA of –5.4 (IQR, –7.4 to –1.7) days based on the POF and CLC, and ΔGA of –3.2 (IQR, –6.8 to 0.3) days based on the CB ( P < 0001 for all). There was no significant difference in brain maturation between fetuses with brain‐sparing and those with non‐brain‐sparing FGR or between male and female fetuses. The EFW percentile correlated significantly with the degree of delayed maturation (whole‐scan data, r = 0.377; P = 0.013) and contributed to our multivariable model. Socioeconomic status, fetal sex and CPR were not associated with the delay in maturation. Seventeen neonates were born < 32.0 weeks' gestation and admitted to the neonatal intensive care unit. The ΔGA was higher in the neonates with perinatal complications, indicating greater delay in brain maturation. Conclusions Delayed fetal brain maturation in early‐onset FGR has been demonstrated using a 3D ultrasound deep‐learning model. These findings highlight the potential role of 3D ultrasound in the assessment of fetal brain maturation and support the need for continued research into these findings and their long‐term neurodevelopmental consequences in early‐onset FGR. © 2026 The Author(s). Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.