Reducing Stillbirths with Smartphone Technology

Stillbirth pregnancy loss after 28 weeks occurs every 16 seconds globally. 

Uniting a leading maternal-fetal medicine program with some of the world’s most advanced supercomputing resources, researchers and clinicians at UT Austin are aiming to reduce rates of stillbirth, which occurs more than 20,000 times per year in the U.S. alone.

These findings are especially exciting because fetal breathing and hiccups are strong predictors of a normal fetal outcome.

Monitoring a fetus’ movements can be the most helpful health indicator in late pregnancy, but traditional, manual methods like having a mother track “kick counts” can be limited to as little as 30 percent overall accuracy.

The solution may already be in your pocket.

Kenneth Moise, a maternal-fetal specialist and professor at Dell Medical School, has launched a study in collaboration with Kelly Gaither, the deputy director at TACC and an associate professor in the Department of Women’s Health at Dell Med, to create a deep-learning algorithm that relies on smartphone audio recordings to track fetal movements.

“By identifying deviations from normal fetal movement patterns, we can intervene before a crisis arises, and more broadly, change the landscape of later-stage prenatal care,” says Moise, who also leads the Comprehensive Fetal Care Center at Dell Children’s Medical Center. 

Pilot studies of the algorithm indicate an accuracy rate of more than 70 percent in detecting overall fetal movement, already more than doubling the accuracy of traditional methods and creating a pathway for consistent, reliable, at-home monitoring.

With the power of machine learning and artificial intelligence, we can build an algorithm to reduce the guesswork and alleviate unnecessary anxiety for the patient while providing critically needed data to inform clinicians.

Pilot studies show the algorithm outperforms maternal perception by a wide margin — detecting overall fetal movement with over 87 percent accuracy, more than 4.5 times the maternal detection rate. It also identified 71 percent of fetal breathing movements (compared to just 3 percent perceived by patients) and 92 percent of fetal hiccups (versus 32 percent perceived by patients). “These findings are especially exciting,” Moise said, “because fetal breathing and hiccups are strong predictors of a normal fetal outcome.”

The data gives providers an additional tool to intervene earlier if concerns arise, and in addition to providing real-time data, future applications like a smartphone app could allow mothers to track movement patterns over time, offering a daily snapshot of fetal well-being.

Building the Solution

The current study follows 30 pregnant women over the course of five visits. At each visit, the team tracks three monitoring methods simultaneously: Ultrasound provides high-accuracy control data, while mothers also hold a smartphone to their abdomen to record sound, and they separately track “kicks” in the traditional method.

“The process was so simple, and it’s such an easy thing to do to help progress really valuable technology,” says Sara Toynbee, who participated in pilot studies and returned to participate during her second pregnancy. 

With vast data to reconcile across three input types, TACC’s unique expertise and technical capability ensures a reliable, usable algorithm.

“With the power of machine learning and artificial intelligence, we can build an algorithm to reduce the guesswork and alleviate unnecessary anxiety for the patient while providing critically needed data to inform clinicians,” said Gaither.

Translation to the Real World

In the spring of 2025, the project was submitted for consideration for funding by the Texas Health Catalyst Award program at the Dell Medical School, UT Austin.

“We are proud to be a finalist in this competition and awarded start-up funding to further accelerate the development of our concept,” says Moise.

Gaither and Moise plan to complete two focus groups, gathering insights to shape the app’s design. Next, they’ll run a cross-sectional study, analyzing the results to fine-tune the AI, before aiming to launch a pilot in early 2026.

‘From here, we could release a basic app for parents or pursue multicenter trials for FDA approval, making it a physician-prescribed diagnostic tool,” Moise says. 

“Programs like Texas Health Catalyst help our teams think through the best paths to market — the paths that will have the best patient uptake and impact on outcomes, which is really what we’re after.”