Bold claim: Artificial intelligence can dramatically improve the detection of a dangerous pregnancy complication that often goes unnoticed with standard screening. A newly developed AI model accurately identified placenta accreta spectrum (PAS), a condition where the placenta abnormally attaches to the uterine wall and can lead to severe bleeding and other serious outcomes. This finding was shared today at the Society for Maternal-Fetal Medicine (SMFM) 2026 Pregnancy Meeting. PAS is a leading cause of maternal morbidity and mortality, and current data suggest only about half of cases are diagnosed during pregnancy.
PAS typically occurs when the placenta attaches to the uterine wall in an abnormal way, frequently after prior uterine surgeries such as cesarean delivery. The rate of PAS is rising in the United States. When PAS goes undiagnosed before delivery, it can trigger massive hemorrhage, multisystem organ failure, and even death. Pregnancies with high PAS risk are usually screened using risk factors and ultrasounds to anticipate and manage potential problems before birth, but findings can be inconclusive or misleading.
“We’re very excited about what this AI model could mean for clinical practice, offering accurate and timely PAS diagnosis,” said Alexandra L. Hammerquist, MD, a maternal-fetal medicine fellow at Baylor College of Medicine in Houston, TX. “We hope using it as a screening tool could reduce PAS-related maternal morbidity and mortality.”
In a retrospective study, researchers from Baylor College of Medicine analyzed 2D obstetric ultrasound images from 113 at‑risk PAS patients who delivered at Texas Children’s Hospital between 2018 and 2025. The average gestational age at the maternal ultrasound was about 30.9 weeks (plus or minus 3.7 weeks).
Their AI program demonstrated perfect detection of PAS across these cases: all PAS instances were correctly identified, with only two false positives and no false negatives for placenta accreta.
In short, this AI approach showed promise for enhancing PAS screening accuracy, potentially enabling earlier intervention and better outcomes for mothers and babies.
Controversy-worthy note: While the results are compelling, questions remain about how the model performs in broader, real-world settings and across diverse populations. How might such a tool be integrated into standard prenatal care, and what safeguards are needed to prevent overdiagnosis or unnecessary interventions? Share your thoughts and experiences in the comments.
