The World Health Organization estimates that more than 250,000 preventable deaths occur worldwide each year due to pregnancy-related complications. Prenatal ultrasound can improve outcomes and reduce fetal and maternal mortality, yet two-thirds of people globally lack access to diagnostic imaging like ultrasound. Current methods rely heavily on specialized clinical training, leaving millions in low-resource settings without a single ultrasound during their pregnancy.
To address this gap, researchers developed machine learning systems that interpret "blind sweep" ultrasounds. This scanning method allows frontline health workers without sonography training to perform scans by gliding low-cost, portable probes across the abdomen in a standardized pattern. Because many regions face critical shortages of experts, and traditional sonographer training is expensive and takes years, minimal training combined with AI interpretation could significantly expand prenatal screening coverage.
In a study published in JAMA Network Open, researchers evaluated how well an AI model for gestational age assessment generalizes across diverse clinical settings and hardware. Gestational age, or how far along a pregnancy is, underpins much of prenatal care, including screening, treatment and delivery decisions. Earlier research validated the model on unseen patients from the same cohort it learned from. This prospective study tested whether the system performs reliably running live on mobile devices in real clinics, with novice operators and unfamiliar equipment.
The model was tested at Northwestern Medicine in Chicago, an urban, high-resource academic center, and at Jacaranda Health in Nairobi, Kenya, an urban community clinic network. The sites differed in when women begin prenatal care, how many routine ultrasounds they receive, and how gestational age is dated. Fine-tuning used a small pilot dataset of 120 Chicago participants, about 6% of the original training set, to adapt the model to new probe hardware. No Nairobi data was used, keeping the Kenyan validation a strictly independent test of generalizability.
Source: Google AI Blog · Summarized by HeadlinesBriefing