Redesigning Ultrasound for the Frontlines of Pregnancy Care
When a clinician has to make decisions about a pregnancy without knowing how far along it truly is, every choice becomes harder — and riskier. Early in his career working in Zambia, Dr. Jeffrey Stringer saw this reality play out every day, as the absence of ultrasound forced providers to care for women with critical information missing. “I saw firsthand how the absence of ultrasound undermined our ability to provide good obstetric care,” he said.
Stringer, now a physician-researcher at the University of North Carolina at Chapel Hill, spent more than a decade working in settings where access to prenatal imaging was limited or nonexistent. Without ultrasound, clinicians were left to estimate gestational age, assess fetal growth, and rule out complications using incomplete information — decisions that shape outcomes for both mothers and babies.
The full scope of the problem became clear through research. While analyzing data from an ongoing clinical trial, Stringer and his team noticed what appeared to be an implausibly high preterm birth rate. The numbers raised immediate concern. After deeper analysis, they realized the issue was not that women were delivering early, but that pregnancies were being dated inaccurately. “We realized the root cause was not that the preterm birth rate was that high,” Stringer explained, “but that it only seemed high because inaccurate gestational dating due to lack of ultrasound.” That single gap in information propagated across nearly every aspect of care.
Modern prenatal medicine depends on timing. Screenings, interventions, and clinical decisions follow a carefully sequenced schedule tied directly to gestational age. “If gestational age is wrong, care is mistimed,” Stringer said, “which directly compromises maternal and fetal outcomes.” Without ultrasound, clinicians are forced to make high-stakes decisions without the clarity that modern care assumes.
Those early experiences also shaped how Stringer thinks about maternity care deserts in the United States. While often described as a problem of distance, access is more frequently constrained by logistics. “Inability to leave work, arrange childcare, or afford transportation mean that care is delayed or never received,” he said. The downstream effects mirror what he witnessed internationally: delayed diagnoses, missed intervention windows, and preventable harm.
Rather than attempting to adapt traditional ultrasound systems to these environments, Stringer and his collaborators began with a different premise: ultrasound should be designed for the realities of frontline care. “Our global health experience defined the design requirements,” he said. “We built the system for environments with limited training, constrained infrastructure, and high clinical need.”
It’s called the FAMLI app—and at the core of the system is a fundamentally different way of collecting imaging data. Instead of requiring a clinician to know exactly what anatomical plane to find, the approach relies on standardized “blind sweeps” — short video clips captured as a probe moves methodically across the abdomen. “You’re collecting an ultrasound without needing to know exactly what you’re looking at,” said Katelyn Rittenhouse, MD, NCInnovation Clinical Lead, Assistant Professor. “That’s the unique thing. We replaced expert image acquisition with a reproducible process.”
Those standardized sweeps — now numbering more than 50,000 — form the foundation of the AI models. Rather than analyzing isolated images, the models ingest the data end-to-end, learning directly from the full sweep videos. “With AI, these models can analyze images in ways that move beyond how ultrasound has traditionally been done,” Teeranan “Ben” Pokaprakarn, NCInnovation Machine Learning Scientist, Research Assistant Professor, explained. The system can estimate gestational age and identify pregnancy-related complications without requiring interpretive decisions from the user.
Building those models required significant institutional investment. Training took place on a large cluster of GPUs (graphic processing unit) housed on campus, supported by university resources and supplemented by the Gates Foundation and the National Institutes of Health. But making the technology usable in real-world settings required another leap: miniaturization. “We had to take these powerful models and make them small enough to run on a mobile device,” said Srihari “Sri” Chari, Project Manager. The result is a system that can run entirely offline — no server, no internet connection — using only a tablet and a handheld ultrasound probe.
That capability matters in both global and domestic contexts. In eastern North Carolina’s rural communities, clinics often lack on-site obstetric ultrasound despite seeing pregnant patients every day. The team has begun clinical deployments in affiliated UNC clinics and hospitals, including sites in Raleigh and Wilmington, with discussions underway in additional counties. In every setting, the system operates alongside standard clinical workflows, with conventional ultrasound used as a backup for comparison and validation.
Parallel to the technical work, the team has undertaken extensive customer discovery to ensure the technology fits how care is actually delivered. “What works in Zambia isn’t automatically what works in North Carolina,” said Harmony Chi, Program Manager.
Through more than 20 discovery calls with rural health clinics, family medicine practices, and frontline providers, the team identified the diagnoses that matter most in domestic settings — including early pregnancy viability, ectopic pregnancy, and growth monitoring.
Those conversations revealed a consistent theme: clinics are overwhelmed. Sonographers are in short supply. Family medicine providers often lack confidence in making obstetric diagnoses without imaging. Routine scans consume time and resources, while patients travel long distances for evaluations that frequently rule out serious concerns. “The goal is to push ultrasound upstream,” Dr. Stringer said, “so anyone can do it, anywhere.”
The potential system-level benefits are significant. Growth monitoring alone can require hours of scanning every two weeks, even for low-risk patients. By ruling out normal cases earlier, clinics could reduce unnecessary referrals, save healthcare dollars, and spare patients long drives to tertiary centers. “There are upstream benefits for both the system and the patient,” Dr. Rittenhouse noted.
NCInnovation’s support has been instrumental in moving this work from research toward real-world deployment. Through its Entrepreneur in Residence program, NCInnovation has connected the team to commercialization expertise, regulatory consultants, and partners across the innovation ecosystem. “They understand both the science and what it takes to bring something to market,” said Chari.
The team is now preparing to submit a formal application to the FDA. Multiple commercialization pathways are under evaluation, including distribution through existing ultrasound platforms.
For Stringer, the work remains deeply personal. He speaks openly about how access to ultrasound shaped his own family’s experience — and how its absence can change outcomes in ways that are both profound and preventable. Success, for him, is not defined by technology alone. It means seeing a viable product in the field, improving pregnancy care where access has long been limited, and ensuring that innovations originating at UNC translate into real impact for women in North Carolina and beyond.
