Delivery Date AI is an FDA De Novo Granted, cloud-based Software as a Medical Device (SaMD) that analyzes raw ultrasound images to provide a Predicted Delivery Date (PDD) for patients without a reliable estimated delivery date. Using an ensemble of deep-learning neural networks, the system processes entire ultrasound images — capturing both fetal and maternal characteristics — to generate an independent, image-based date estimate. It is designed for use in singleton pregnancies between 14 0/7 and 36 6/7 weeks gestation where standard dating methods are unavailable or unreliable. The Predicted Delivery Date is designed to support your assessment, not replace it. It works alongside your clinical judgment and standard-of-care methods — an independent data point when you need one most.

When a patient arrives without a reliable last menstrual period, no first-trimester ultrasound, or late into pregnancy, you're dating without a foundation. Standard biometric methods assume fetal size reflects maturity — which can systematically misclassify small-for-gestational-age fetuses as younger than they are, compounding the uncertainty in exactly the cases where accuracy matters most.

Delivery Date AI analyzes whole images rather than isolated measurements, drawing on characteristics that are biologically independent of fetal growth. The result is an independent estimate your team can use to inform decisions with more confidence.

The PDD is an adjunctive tool. It is intended to aid clinical judgment alongside your standard-of-care methods for assessing gestational age — not to serve as the sole basis for clinical decisions.

When the PDD and your biometric EDD are in agreement (within 10 days), it provides evidence-based confidence in your current dating. When they differ by more than 10 days, that discrepancy is a signal to reassess the full clinical picture — patient history, growth trajectory, and the possibility of fetal growth deviation — before making time-sensitive management decisions.

The 10-day threshold is anchored to the biological precision limits of biometric dating established in foundational literature (Benson, Hadlock), which recognizes ultrasound dating accuracy in mid-pregnancy at approximately ±7–10 days. Delivery Date AI holds that same conservative threshold across the full indicated gestational age range.

In the pivotal clinical validation study — 247 suboptimally dated singleton pregnancies between 14 0/7 and 36 6/7 weeks gestation — Delivery Date AI demonstrated statistically significant superiority over the standard-of-care Hadlock method.

The AI achieved a Mean Absolute Error (MAE) of 15.22 days (95% CI: 13.67–16.77), compared to Hadlock's MAE of 36.41 days (95% CI: 32.88–39.94). That represents a 21.19-day absolute improvement over the pre-specified 7-day superiority margin (p<0.001) — a 58% reduction in prediction error compared to current standard of care.

These results apply specifically to the suboptimally dated population for which the device is indicated and should not be generalized to all pregnancies.

Yes. The PAIR (Perinatal Artificial Intelligence in ultRasound) study, published in the Journal of Maternal-Fetal & Neonatal Medicine (Patel et al., 2025), evaluated the AI platform's delivery timing performance across a much larger and more diverse dataset.
The PAIR study enrolled 5,714 patients across 19,940 ultrasound examinations at the University of Kentucky and satellite clinics, using over 2 million de-identified images. Across all births, the most current model version (V4) achieved an R² of 0.92 for delivery timing correlation, with an MAE of 12.90 days — and R² of 0.95 for term births specifically. Prediction accuracy was preserved across all three trimesters, with MAE ranging from 10.76 to 15.06 days.

The PAIR study also demonstrated that retraining the model with additional data progressively improved performance, highlighting the platform's continuous learning architecture.

The FDA pivotal validation cohort (N=247) included women aged 18 years and older with singleton pregnancies between 14 0/7 and 36 6/7 weeks gestation who lacked a reliable EDD — defined as an unreliable or unknown LMP and no first-trimester dating scan. The cohort was 62.8% obese (BMI ≥30) and included 63.2% White and 31.8% Black or African American patients. Asian and Hispanic populations were under-represented; results may be less characterized for these groups.

The PAIR study (N=5,714) included all pregnancies — term and preterm — across a broader population, providing a wider view of the underlying model's delivery timing capabilities across gestational ages and clinical scenarios.

Delivery Date AI has been validated for use with 2D grayscale (including Doppler) images acquired from GE and Philips ultrasound systems, using convex, micro-convex, and transvaginal probes. Phased array and linear probes are not supported.

The system accepts DICOM files (.dcm) containing standard-of-care 2D grayscale static images, including Doppler. JPEG, PNG, 3D/4D, M-mode, cine loops, and video formats are not supported and will be automatically rejected. A minimum of more than 10 valid images is required for the system to generate a prediction.

The device is not intended for use in multiple gestations, pregnancies with known fetal anomalies, pregnancies under 14 0/7 weeks gestation, or pregnancies at or beyond 37 0/7 weeks gestation. It does not predict the risk of preterm birth, and it does not forecast medically induced delivery — the validation population comprised pregnancies with spontaneous onset of labor only.

Using the device outside these parameters is off-label and outside the validated performance envelope.

Yes. Delivery Date AI can be used on standard-of-care ultrasound examinations throughout the indicated gestational age window (14 0/7 – 36 6/7 weeks), consistent with the indications for use and your clinical judgment. Use should always be in the context of an eligible patient — a suboptimally dated singleton pregnancy where a reliable EDD is unavailable.

Delivery Date AI has received FDA De Novo designation, establishing a new regulatory classification for AI-powered delivery date prediction software (Class II SaMD, Product Code SHE, 21 CFR 892.8200). It is also approved in Brazil by ANVISA. Federal law restricts this device to sale by or on the order of a physician or other licensed healthcare professional.

Delivery Date AI is designed for use by board-certified or board-eligible OB/GYNs and Maternal-Fetal Medicine (MFM) specialists. It is a prescription medical device intended exclusively for licensed healthcare practitioners.

Delivery Date AI complies with HIPAA and relevant privacy regulations. All data transmitted between the client and server is encrypted using SSL/TLS protocols. The platform supports two-factor authentication for enhanced account security. Patient names are displayed using only the first two characters of first and last name to minimize identifiable information within the interface.