Abstract

Introduction: Obstetric ultrasound is fundamental in prenatal care for gestational age (GA) estimation, fetal monitoring, and complication screening. However, access to quality ultrasound is limited in many low- and middle-income countries (LMICs), where nearly half of pregnant women receive no scans during pregnancy. Even in high-income countries, disparities in care persist. Recently, artificial intelligence (AI) applied to "blind" ultrasound sweeps-standardized transabdominal sweeps performed by minimally trained personnel-has emerged as a promising tool to improve access to diagnostic-quality prenatal ultrasound.

Material and methods: A systematic review following PRISMA guidelines was conducted. PubMed was searched through April 2025 using terms such as [blind sweep], [prenatal ultrasound], and [deep learning]. Studies were included if they assessed AI models applied to blind-sweep ultrasound for prenatal diagnostics. Fourteen studies (12 original, 2 reviews/meta-analyses) met eligibility. Data were extracted on study design, population, acquisition protocol, AI models, and diagnostic performance. Risk of bias was assessed using QUADAS-2.

Results: AI models demonstrated comparable or superior performance to expert sonographers in mid-trimester GA estimation, with mean absolute errors of 3-5 days. In a large multicenter study, AI outperformed traditional biometry (3.9 vs. 4.7 days error). Accuracy remained high even with minimally trained operators. AI also performed well in detecting breech presentation (AUC ~0.98), assessing amniotic fluid (Dice ~0.88; AFI accuracy ~91%), and segmenting fetal anatomy for biometry. Limitations included reduced accuracy in late pregnancy and limited validation in early gestation or anomaly detection.

Conclusions: AI-based blind sweep ultrasound holds transformative potential for global prenatal care, enabling scalable, low-cost diagnostics in LMICs and underserved settings. While promising, clinical adoption requires broader validation, improved interpretability, and integration into healthcare systems. With further development, this technology could significantly contribute to equitable prenatal diagnostics and reduced maternal-fetal morbidity worldwide.