On March 2, 2026, the FDA granted De Novo clearance to Ultrasound AI for its Delivery Date AI technology, a cloud-based Software as a Medical Device (SaMD) designed to predict a delivery date (PDD) using only standard ultrasound images.¹ The platform is intended as an adjunctive tool to support clinical decision-making, particularly in pregnancies where traditional dating methods—such as last menstrual period (LMP) or standard gestational-age ultrasound measurements—are deemed unreliable.

"Fetal development and delivery timing impact nearly every decision in obstetrics," said Nathan Fox, a board-certified MFM specialist and partner physician with Ultrasound AI, in a statement. "Delivery Date AI gives clinicians unprecedented insight into a pregnancy's progression, allowing us to make more informed decisions for how and when we intervene."¹

The clearance is supported by data from the Perinatal Artificial Intelligence in Ultrasound (PAIR) study, published July 27, 2025, in The Journal of Maternal-Fetal & Neonatal Medicine.² Conducted in collaboration with the University of Kentucky, the study evaluated the AI’s ability to forecast delivery timing by analyzing fetal and maternal characteristics within a vast dataset of de-identified images.

“This [clearance], to me, is super exciting, because, for one, we don't have a lot of tools that come up in pregnancy,” said Raquel Dardik, MD, Chief medical officer, Ultrasound AI; Clinical associate professor, OBGYN, Herbert Wertheim College of Medicine, in a video interview with Contemporary OB/GYN. “Oftentimes, we have something that can slightly improve how this looks or can slightly improve how this gets managed, but it is unusual to have something that can actually change the way we look at things on a clinical basis,” added Dardik.

Study design and data evolution
The PAIR study utilized a massive longitudinal dataset to train and validate the ensemble of deep-learning neural networks.² The initial training phase involved 5,714 pregnant women, encompassing 19,940 unique ultrasound exams and 877,141 total images. To assess the model's accuracy, a validation subgroup of 1,209 patients was reserved, with delivery outcomes blinded from the AI by an independent third-party monitor.

Performance was measured through multiple iterations of the software. Version 1 (V1) was initially trained on the 2017–2021 cohort, while subsequent versions (V3 and V4) incorporated an additional 1,165,618 images through 2023.² This iterative retraining demonstrated a consistent upward trajectory in predictive accuracy. The R2 value for all births increased from 0.85 in V1 to 0.88 in V3, ultimately reaching 0.92 in the final V4 model.

Predictive accuracy and preterm birth insights
A primary objective of the PAIR study was to assess the AI’s performance in predicting spontaneous preterm birth (PTB).² In the initial model, the AI exhibited a sensitivity of 39% and a specificity of 93% for PTB prediction, with an area under the curve (AUC) of 0.757. While the R2 for term births was high at 0.90, the R2 for spontaneous PTB alone was 0.48 in the early version, eventually achieving a mean absolute error (MAE) of 19.99 days in the V4 iteration.

Crucially, the study found that the MAE in predicting the number of days until delivery remained consistent across all trimesters assessed.² This "image-first" approach allows the technology to derive insights from entire ultrasound images rather than relying on manual biometry, which can be subject to inter-operator variability.

Clinical integration and workflow
The Delivery Date AI is designed to integrate into existing obstetrics and maternal-fetal medicine (MFM) workflows without requiring new hardware.¹ The system is compatible with most existing ultrasound machines, uploading images to the cloud and delivering results within seconds.

By providing a real-time Predicted Delivery Date, the technology aims to reduce the clinical uncertainty that often complicates the management of high-risk pregnancies or those with late-entry prenatal care. Robert Bunn, President and Founder of Ultrasound AI, noted that the clearance represents a milestone in reducing the burden of preterm birth by supporting earlier clinical decision-making.

“Delivery Date AI isn't just innovative; it's a clinically evaluated tool that helps clinicians reduce uncertainty and better support mothers and families,” said Bunn in a press release.

The developer suggests the tool is particularly scalable for resource-constrained clinics and "obstetric deserts," where access to advanced dating expertise may be limited.