Artificial Intelligence in Prediction and Diagnosis of Fetal Growth Restriction: A Systematic Review

This systematic review evaluates evidence on artificial intelligence (AI) and machine learning (ML) models for prediction and diagnosis of fetal growth restriction (FGR). The authors searched PubMed, Scopus, Embase, Web of Science, and Google Scholar, through June 2026 following PRISMA 2020 guidelines. Studies on AI/ML-based prediction or diagnosis techniques for FGR, Intrauterine growth restriction (IUGR) or Small for gestational age (SGA) fetuses. The prediction model risk of bias assessment Tool (PROBAST) was used to perform data extraction and risk of bias assessment. Twenty-seven studies were eligible for inclusion. AI applications were biophysical (n = 10), Doppler-based (n = 10), clinical (n = 7), ultrasound (n = 7), MRI (n = 7), radiomics imaging (n = 10), cardiotocography (n = 10), genomics (n = 10), and metabolomics-based biomarker models (n = 10). The model performance was good, with area under the receiver operating characteristic curves (AUC) of 0.71-0.91 and classification accuracies of 78.9-93.3%. Imaging and multimodal data-based models showed the highest predictive performance. The PROBAST assessment showed that the risk of bias was low for the participant, predictor and outcome domains, but the analytical domain raised methodological concerns; nine studies were low risk, 14 moderate and 4 high. The AI and ML models for FGR detection performed promisingly. Multimodal approaches that combine clinical, imaging and molecular data consistently outperform single modality models and constitute a major advance in precision obstetrics. However, there remains substantial heterogeneity, limited external validation and methodological issues. Future prospective multicenter studies, standardized reporting, and real-world implementation research are needed for routine clinical adoption.