Google researchers use AI to enable blind-sweep ultrasounds for prenatal care

Google researchers, in partnership with Northwestern and Jacaranda Health, developed an AI tool to address the global shortage of sonographers and diagnostic imaging access. The study focused on enabling healthcare workers to perform simple "blind sweep" ultrasounds, which the AI then interprets to provide critical prenatal data.

How the Technology Works

Instead of requiring precise probe manipulation by a specialized sonographer, operators sweep the ultrasound probe over the abdomen in a predefined pattern. Healthcare workers with no prior ultrasound experience can learn this technique in just eight hours. The AI model analyzes the resulting videos to:

• Estimate gestational age
• Determine fetal presentation (position)
• Instruct operators if a sweep needs to be redone

Crucially, all processing occurs on the device itself, meaning the system works offline without needing stable electricity or Wi-Fi.

Study Results and Impact

The research involved 1,000 mothers in Nairobi, Kenya, and 1,000 in Chicago. The team found that their machine learning model detected gestational age and fetal presentation with accuracy comparable to that of a trained sonographer. This approach offers a scalable solution for under-resourced areas, where traditional ultrasound machines are often too bulky, expensive, or difficult to maintain. By putting expert-level diagnostic capabilities into the hands of frontline community health workers, the technology aims to improve maternal health outcomes globally.