AI enhances prenatal ultrasound access in low-resource settings
In brief
- An AI model interprets prenatal ultrasounds performed by non-specialists using portable probes.
- Accuracy matches experts (4.18 vs 4.48 days mean error) across two distinct clinics.
- Tested without local tuning in Nairobi, demonstrating strong generalization.
Google Research has published a prospective study in *JAMA Network Open* showing that an AI system can interpret prenatal ultrasounds performed by non-specialist operators using portable probes. According to the article, the tool estimates gestational age with accuracy comparable to expert sonographers, even under varied material and geographical conditions.
The study tested the model on 385 participants across two clinics (Chicago and Nairobi), with no local tuning for the latter. Results show a mean absolute error of 4.18 days for AI, versus 4.48 days for sonographers, maintaining robustness in complex scenarios like multiple pregnancies or large fetuses. Researchers highlight the model’s generalization capability across different equipment and populations.
This breakthrough could reduce the 250,000 annual deaths from pregnancy complications by expanding ultrasound access, currently limited by a lack of experts and expensive equipment. The authors stress the need to build trust among clinicians and patients, including integrating confidence indicators to flag poor-quality scans.
What it changes for you
- HealthThe tool could reduce deaths from pregnancy complications by expanding ultrasound access in underserved areas.
This summary gives you the essentials. The full article, its figures and quotes are on Google Research.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error