Mount Sinai's AI Breakthrough: Could AI Imaging Spot Pregnancy Health Risks Earlier?

The Future of Maternal Care: Identifying Risks Sooner

Imagine a world where significant pregnancy complications could be predicted and addressed much earlier, even before conception. This future is rapidly becoming a reality, thanks to pioneering work at Mount Sinai, one of the nation's leading teaching hospitals. They are leveraging cutting-edge artificial intelligence (AI) to transform how we assess pregnancy risks, moving identification upstream in the care pathway.

This innovative approach zeroes in on two critical conditions: Placenta Accreta Spectrum (PAS) and congenital heart defects (CHD). Both are associated with significant health challenges for mother and baby, substantial medical costs, and intensive resource needs. Early identification opens doors to crucial benefits like timely counseling, targeted surveillance, and planned deliveries in specialized facilities, ultimately leading to better outcomes.

Understanding Placenta Accreta Spectrum (PAS) and Congenital Heart Defects (CHD)

Placenta Accreta Spectrum (PAS)is a severe complication where the placenta grows too deeply into the uterine wall. This makes delivery incredibly high-risk and resource-intensive, often requiring complex surgical intervention. Similarly,congenital heart defects (CHD)are structural problems in a baby's heart that develop before birth, demanding specialized care.

AI-Powered Insights: Revolutionizing Pregnancy Risk Assessment

Mount Sinai specialists recently unveiled groundbreaking AI-assisted workflows at the 2026 SMFM Annual Pregnancy Meeting. Their research highlights a dual strategy: using machine learning models to predict PAS risk from preconception data and deploying AI to detect severe CHD during routine mid-trimester fetal ultrasounds. This work signals a significant leap toward data-informed pregnancy care, integrating clinical, social, and operational signals.

Predicting PAS Risk Before Conception

In a comprehensive case-control study involving nearly 119,000 deliveries over a decade (2013-2023), PAS occurred in 0.23% of cases. Despite its rarity, the condition carries high risks for severe maternal morbidity and mortality. This underscores the strategic value of precise preconception risk stratification.

The Mount Sinai team trained multiple machine learning models using extensive pre-pregnancy electronic medical record (EMR) data, including demographics, obstetric and surgical history, vital signs, and lab results. Remarkably, their AI discovered that pre-pregnancy anemia, a factor not previously recognized, is an additional risk for PAS. This finding sits alongside established risks such as older maternal age, a history of C-sections, prior gynecologic surgery, and past pregnancy complications.

The implications are profound. Since anemia is a potentially modifiable condition, healthcare systems could intervene by guiding patients toward nutritional support, consultations, or preconception counseling. The goal is to mitigate risks, avoid emergency deliveries, and enable planned, specialized care.

Technically, an XGBoost model demonstrated superior performance in distinguishing high- from low-risk cases with an AUC of 0.86, outperforming logistic regression (0.76). While a Random Forest model achieved the highest sensitivity at 91% (catching more true cases), logistic regression offered 91% specificity (fewer false alarms). These findings highlight the strategic trade-offs organizations face when optimizing AI models for different clinical priorities.

Detecting Congenital Heart Defects with AI Imaging

On the imaging front, Mount Sinai West has implemented BrightHeart software to enhance fetal ultrasound screening for major CHD. In a study spanning 11 medical centers across two countries, AI assistance significantly boosted the detection of major CHD to over 97%. Beyond accuracy, the technology cut reading time by 18% and increased reader confidence by 19%. This demonstrates clear value for both general gynecologists and maternal-fetal medicine subspecialists reviewing these crucial second-trimester ultrasounds.

This technology is currently undergoing real-world evaluation in a prenatal diagnostic center, where it seamlessly flags suspicious findings for severe CHD within standard screening workflows.Could AI imaging spot pregnancy health risks earlier?For congenital heart defects, the answer is a resounding yes, offering unprecedented precision and efficiency.

Responsible Innovation: Ensuring Ethical AI in Healthcare

The rollout of such advanced AI tools is not without its challenges. Mount Sinai rightly emphasizes the importance of rigorous validation across diverse populations, careful stewardship of large retrospective datasets, and continuous monitoring for potential biases. This includes integrating social indices like composite social vulnerability scores and neighborhood gun violence exposure, ensuring equitable and just application of AI.

The institution also champions the need for strong clinical sponsorship, with measurable metrics tied to morbidity, cost reduction, and workflow improvement. A deliberate, scalable plan, transitioning from single-center pilots to system-wide decision support, is crucial for successful integration.

Redefining Pregnancy Care Through AI

By synergizing EMR-driven preconception risk prediction for PAS with AI-augmented fetal cardiac imaging, Mount Sinai is fundamentally redefining when and how pregnancy risks are identified. The early results point to tangible gains in diagnostic accuracy, operational efficiency, and proactive care planning. These advancements hold the promise of a healthier future for mothers and babies, provided health systems thoughtfully match technical performance with robust governance and seamless integration into clinical workflows.