AI for Pregnancy: New Tool Promises to Predict Due Dates From Ultrasounds Alone

When you’re expecting a baby, one of the first questions everyone asks is simple: When is the due date? But determining that date isn’t always straightforward. For some pregnancies, the usual clues — like the first day of a last menstrual period or early ultrasound measurements — can be unclear or unavailable. Now, a new artificial intelligence tool aims to help fill that gap.

Ultrasound AI, a company focused on applying machine learning to medical imaging, has received FDA De Novo clearance for Delivery Date AI, a cloud-based technology designed to estimate a baby’s predicted delivery date using only standard ultrasound images.

The system analyzes ultrasound scans and provides clinicians with a real-time predicted delivery date during routine prenatal visits, potentially helping doctors make more informed decisions about pregnancy care.

The technology was trained on millions of de-identified ultrasound images from pregnancies across different clinical settings. Using deep-learning neural networks, the system examines features within the ultrasound images — including both fetal and maternal characteristics associated with delivery timing — to estimate how far along a pregnancy is and forecast when delivery is likely to occur.

According to the company, the tool is intended for situations where traditional dating methods may not be reliable. In those cases, turning imaging data into a predicted delivery timeline could give clinicians another data point when planning care.

“Delivery Date AI represents a genuine technological advance — it is the first FDA-cleared tool to predict delivery dates from ultrasound images alone, and it fills a real clinical gap for patients lacking reliable traditional dating methods,” says Dr. Zaid Fadul, CEO of Bespoke Concierge MD, who is not affiliated with Ultrasound AI. “Its potential to improve care in resource-constrained settings and obstetric deserts is meaningful.”

In a peer-reviewed study involving more than 5,700 patients, researchers reported that the system showed strong accuracy when predicting time to delivery for pregnancies that went to full term. The study found that the model achieved a high statistical correlation when estimating days to delivery using ultrasound images alone.

For clinicians, the appeal of a tool like this lies in its simplicity. Delivery Date AI can integrate with existing ultrasound machines and prenatal workflows, producing predictions in seconds once images are uploaded. That ease of use could make it particularly useful in clinics with limited resources or in areas where access to specialized prenatal care is more limited.

Still, experts say that the technology should be viewed as a supplement to clinical judgment rather than a replacement for it.

“AI ultrasound tools like Delivery Date AI work best at estimating how far along a pregnancy is during the middle months (roughly 14 to 28 weeks) where studies show they can predict gestational age within about 3 to 5 days, which is close to what an experienced ultrasound technician can do,” Fadul says. “The tool can also predict a likely delivery date for pregnancies that end up going full term, with strong accuracy.”

But he notes that the technology has limitations, particularly when it comes to predicting complications.

“It does not predict complications, does not assess preterm birth risk (per the FDA), and performs significantly less well for the very pregnancies where accurate timing predictions matter most,” Fadul says.

One reason is that many AI systems in obstetrics are trained on relatively narrow datasets, he says. If those datasets don’t reflect the diversity of real-world patients, predictions may not perform equally across all populations.

There is also the human factor. As AI tools become more common in healthcare, Fadul warns that physicians must guard against relying too heavily on automated results.

“There’s also a real risk of overreliance,’ where doctors may start trusting the AI’s output too much and pay less attention to their own clinical instincts,” Fadul says. “AI should be treated as one helpful piece of the puzzle, not the whole picture, and every prediction it makes should lead to a conversation with your doctor — not replace one.”

For expecting parents, that distinction matters. AI may be able to analyze images faster than ever before, but pregnancy remains a complex, deeply personal experience shaped by far more than what appears on a scan.