AI-enhanced, portable ultrasound tool reliably measures gestational age

Key takeaways:

• An AI-enhanced, low-cost ultrasound tool performed similarly to standard ultrasound in measuring gestational age for pregnant women.
• The tool could be useful for obstetric care in low-resource settings.

A low-cost, battery-powered ultrasonography probe with integrated AI for image interpretation reliably measured gestational age compared with standard ultrasound, providing a possible option for low-resource settings, diagnostic data show.

WHO recommends all pregnant people receive at least one ultrasonography examination before 24 weeks’ gestation; however, the policy recommendation remains largely aspirational in many low- and middle-income countries, Jeffrey S. A. Stringer, MD, FACOG, professor of obstetrics and gynecology at University of North Carolina School of Medicine and director of UNC Global Women’s Health, wrote in JAMA. Gestational age is used for important care decisions such as when to screen for gestational diabetes or administer certain vaccines, as well as whether to provide corticosteroids or neuroprotective magnesium sulfate for anticipated preterm deliveries.

“Gestational age — and by extension, the due date — is a critical piece of clinical information that is needed for safe management of pregnancy,” Stringer told Healio. “Accurate estimation of gestational age typically requires ultrasound, but ultrasound is not available in many settings. This research demonstrates that a novice provider, armed with a low-cost, AI-enabled ultrasound tool, can estimate gestational age as accurately as an expert sonographer using high-end equipment. Thus, this new technology jumps over two traditional barriers to ultrasound access: expensive equipment and trained sonographers.”

Standard vs. AI-enabled sonography tool

For the prospective diagnostic study, researchers enrolled 400 women with a viable, singleton pregnancy during their first trimester in Lusaka, Zambia, and Chapel Hill, North Carolina. Credentialed sonographers confirmed the women’s due dates with transvaginal ultrasonography, using crown-rump length measurement. At random follow-up visits throughout gestation, including a primary evaluation window from 14 weeks’ to 27 weeks’ gestation, clinicians with no prior ultrasound training obtained “blind sweeps” of the maternal abdomen using the AI-enabled device (index test) and credentialed sonographers performed fetal biometry with a high-specification machine (study standard). The primary outcome was the mean absolute error of the index test and study standard, calculated by comparing each method’s estimate to the previously established gestational age. Gestational ages were considered equivalent if the difference fell within a prespecified margin of 2 days.

The findings were published in JAMA.

Researchers found that the index test with the handheld device met predefined criteria for statistical equivalence to standard ultrasound during the primary evaluation window of 14 to 27 weeks’ gestation. The mean absolute error for the AI-enabled device was 3.2 days vs. 3 days for standard ultrasound, for adifference of 0.2 days (95% CI, 0.1 to 0.5). The proportion of AI-aided assessments correctly classified within 7 days of the gestational age determined by credentialed sonographers were comparable between the two methods, at 90.7% for the index test and 92.5% for the AI-assisted handheld device.

Findings were consistent in analyses stratified by geography and when limited to women with elevated BMI.

“This has major implications for expanding access to ultrasound dating in low-resource settings where trained sonographers and expensive machines are often unavailable,” Stringer told Healio. “We are actively working on new AI models to make additional diagnoses, such as fetal presentation, twins and amniotic fluid volume. Our ultimate goal is to create a suite of tools that can be deployed in low-resource settings to radically improve pregnancy outcomes.”

Need for caution

In a related editorial published in JAMA, Alexis C. Gimovsky, MD, a maternal-fetal medicine specialist and associate professor in the department of obstetrics and gynecology at Brown University, and colleagues wrote that, despite the potential of AI-enhanced ultrasonography, caution is needed for assessments such as anatomic survey and Doppler studies, as the techniques are limited by unpredictability of fetal movement and changes in fetal anatomy over time.

“We caution that although blind sweep ultrasonography dating for gestational age using a low-cost, easy-to-learn technique is a major step forward, it is far from a silver bullet,” Gimovsky and colleagues wrote. “Ultrasonography, whether standard or AI-enabled, will only have the desired impact of improving obstetrics care for the pregnant individual and the newborn when it is implemented within a broader health care system in low- and middle-income countries with trained health care professionals, basic medications and equipment, and a referral system for managing the pregnancy complications identified with the aid of ultrasonography technology.”

As Healio previously reported, data from a single-center, prospective study demonstrated that blind ultrasound sweeps performed by operators without formal training using a low-cost, portable, battery-powered device resulted in high sensitivity and specificity for high-risk pregnancy complications. In that study, published in Obstetrics & Gynecology, sensitivity for detecting a prespecified complication was 91.7% for women with abnormal pregnancy results, with the highest detection rate of 100% for multiple gestations and 91.8% for noncephalic presentation.

References:

• Gimovsky AC, et al. JAMA. 2024;doi:10.1001/jama.2024.14794.
• Toscano M, et al. Obstet Gynecol. 2023;doi:10.1097/AOG.0000000000005139.

For more information:

Jeffrey S. A. Stringer, MD, FACOG, can be reached at jeffrey_stringer@med.unc.edu; X (Twitter): @uncglobalhealth.