Ask a Scientist: How are researchers using AI to help pregnant women access ultrasounds?

Google researchers collaborated with Northwestern Medicine and Jacarand Health to train healthcare workers in performing simple "blind sweep" ultrasounds, enabling AI models to interpret results and provide critical prenatal information such as gestational age and fetal position. This approach addresses the global shortage of sonographers and the lack of access to diagnostic imaging in under-resourced regions, where two-thirds of people lack ready access to such services.

The study involved 1,000 mothers in Nairobi, Kenya, and 1,000 in Chicago, demonstrating that the AI model could detect gestational age and fetal presentation as accurately as trained sonographers. The "blind sweep" method allows operators to gather necessary imaging by sweeping the probe over the abdomen in a predefined pattern, a technique that can be learned in eight hours by workers with no prior ultrasound experience.

The technology operates offline and on-device, eliminating the need for stable electricity or internet connectivity, which are often unavailable in remote clinics. This portability and self-contained processing make it particularly suitable for low-resource settings where traditional ultrasound machines are bulky, expensive, and difficult to maintain.

The research highlights the potential for AI to expand access to expert-level maternal healthcare by empowering frontline community health workers. While the study focused on gestational age and fetal presentation, the team suggests broader applications for ultrasound in other medical contexts, emphasizing the scalability of AI-driven solutions in healthcare.