This whitepaper reviews AI support for screening, imaging and prediction in gynecology and obstetrics, with a focus on Latin America. It is a narrative evidence synthesis built from PubMed searches run in June and July 2026, with a final corpus of 27 references, and it sorts each application into five maturity levels, from retrospective development to monitored routine implementation. It is non-peer-reviewed gray literature published by Arkangel AI, and it does not evaluate any Arkangel product.
Cervical screening and image analysis have the most mature evidence: one cervical cytology model reached an AUC of 0.947 in prospective validation with 2,780 participants, while a liquid-based cytology model showed lower specificity than cytopathologists (0.831 versus 0.901). Fetal ultrasound is an active area, and machine-learning models for preeclampsia and gestational diabetes are numerous but heterogeneous, with pooled AUROC of 0.8492 for gestational diabetes and substantial heterogeneity. Direct regional evidence comes from two Mexican cohorts, including a preeclampsia model with an AUC of 0.897 for preterm disease. The authors conclude that no domain yet supports a claim of safe routine implementation in Latin America and call for external validation, bias audits and process measures before claims of clinical impact.
