AI Matches Sonographers in Estimating Gestational Age

TOPLINE:

Novice users of an ultrasonography tool with AI capabilities were able to estimate gestational age as accurately as credentialed sonographers, new research shows. The mean absolute error was 3.2 days with the AI tool, comparable to 3 days with standard ultrasonography.

METHODOLOGY:

• Researchers conducted a prospective study with 400 pregnant individuals in Lusaka, Zambia, and Chapel Hill, North Carolina.
• Sonographers established gestational age during an index visit.
• During follow-up visits, novice users performed blind sweeps of the maternal abdomen using an AI-enabled device, and credentialed sonographers conducted fetal biometry with high-specification machines to estimate gestational age.
• The primary outcome was the mean absolute error of the AI approach vs standard ultrasonography in the primary evaluation window (14-27 weeks).

TAKEAWAY:

• The mean absolute error was similar for AI and standard ultrasonography, with a difference of 0.2 days (95% CI, −0.1 to 0.5).
• The percentage of gestational age estimates within 7 days of the established gestational age was comparable between the AI tool and standard ultrasonography (90.7% vs 92.5%).
• The AI tool performed consistently across study sites and body mass index categories.

IN PRACTICE:

“These findings have immediate implications for obstetrical care in low-resource settings, advancing the World Health Organization goal of ultrasonography estimation of GA [gestational age] for all pregnant people,” the study authors wrote.

Although this “low-cost, easy-to-learn technique is a major step forward, it is far from a silver bullet,” Alexis C. Gimovsky, MD, with the Women & Infants Hospital of Rhode Island in Providence, Rhode Island, and colleagues wrote in an editorial accompanying the study. Healthcare systems also need “trained healthcare professionals, basic medications and equipment, and a referral system for managing the pregnancy complications identified with the aid of ultrasonography technology,” they noted.

SOURCE:

Jeffrey S.A. Stringer, MD, of the University of North Carolina School of Medicine at Chapel Hill, North Carolina, was the corresponding author on the study, which was published online on August 1 in JAMA.

LIMITATIONS:

The study had a relatively modest sample size and excluded participants with high-risk conditions and known fetal anomalies.

DISCLOSURES:

The Bill & Melinda Gates Foundation funded the research. Butterfly Systems donated ultrasonography probes and collaborated with the investigators to integrate the deep learning model into their device software. One of the editorialists disclosed receiving grants from the Gates Foundation and the National Institutes of Health.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.