A NEW AI-based analysis of fetal growth restriction (FGR) has identified previously unrecognised high-risk scenarios that significantly increase the likelihood of perinatal morbidity and mortality, with risk estimates varying up to tenfold within clinically similar cases.
FGR is a major contributor to stillbirth and neonatal complications, yet predicting which cases will result in adverse outcomes remains challenging. Current clinical guidelines apply similar management strategies across a broad spectrum of FGR cases, despite significant variability in individual risk. To address this, researchers analysed a cohort of 9,558 pregnancies using a probabilistic graphical model (PGM), a form of explainable AI, to quantify how different risk factors interact to influence perinatal outcomes.
The study cohort was divided into an 80% training set (7,645 pregnancies) and a 20% validation set (1,912 pregnancies). The AI model, incorporating the 16 most predictive variables, demonstrated strong performance in distinguishing cases of FGR that resulted in perinatal morbidity (AUC 0.83, 95% CI 0.79–0.87) and remained effective even for rare clinical scenarios (AUC 0.81, 95% CI 0.72–0.90). Notably, the model identified striking differences in risk between seemingly similar cases. For example, female foetuses with an estimated foetal weight (EFW) in the 3rd–9th percentile and no additional risk factors had a perinatal morbidity risk similar to the general cohort (RR 0.9, 95% CI 0.7–1.1), while adding maternal pre-existing diabetes and progesterone use to this scenario increased risk nearly tenfold (RR 9.8, 95% CI 7.5–11.6). A previously unrecognised interaction between foetal sex and maternal diabetes was also uncovered, reversing the typically protective effect of female fetal sex.
These findings underscore the potential for AI-based models to refine risk stratification in obstetric care. The ability to quantify individualised risk and identify latent interactions could support more precise, personalised management of FGR cases, potentially reducing unnecessary interventions for low-risk pregnancies while ensuring targeted monitoring for those at highest risk. Although further validation in diverse populations is needed before clinical implementation, this study highlights the transformative potential of explainable AI in obstetrics. Future work should focus on integrating such models into clinical workflows to enhance decision-making and improve perinatal outcomes.
Katrina Thornber, EMJ
Reference
Zimmerman et al. AI-based analysis of fetal growth restriction in a prospective obstetric cohort quantifies compound risks for perinatal morbidity and mortality and identifies previously unrecognized high risk clinical scenarios. BMC Pregnancy Childbirth.2025;25:80.
