Google’s AI brings prenatal ultrasounds closer

Google tested an artificial intelligence system that makes it possible to perform prenatal ultrasounds with basic training and without relying on an ultrasound specialist. The goal is to bring this care to pregnant people living in areas where equipment, medical staff or a stable internet connection are in short supply.

The research was carried out with Northwestern Medicine and Jacaranda Health with 2,000 women: 1,000 in Nairobi, Kenya, and 1,000 in Chicago. According to results shared by Google, the model was able to estimate gestational age and detect the fetus’s position with accuracy comparable to that of a trained technician.

An ultrasound without having to be a specialist

In a traditional ultrasound, the technician moves the probe across the abdomen in precise positions to obtain different measurements. The procedure requires extensive hands-on training. Google says training a specialist can take around two years.

The study tested another technique called a “blind sweep.” A health worker moves the probe across the abdomen following a predefined pattern, without having to manually locate each part of the fetus. Training to learn this procedure lasted eight hours.

The device captures videos from inside the abdomen, and the AI model analyzes them. It can also tell the operator if the sweep needs to be repeated and provide two important pieces of information:

• Gestational age, meaning how many weeks the pregnancy has lasted.
• Fetal presentation, which indicates the baby’s position.

The analysis takes place directly on the device. That means the system can work without Wi-Fi and without relying on a stable electrical connection while processing the results.

Why knowing how many weeks pregnant someone is matters

Gestational age is not only used to calculate an expected due date. It also helps medical staff decide how to monitor the pregnancy, when to treat a complication and what care to prepare for the birth.

An error of a few weeks can change medical decisions. A baby believed to be 37 weeks along could actually be 34 weeks, a stage when the lungs are less mature and the baby may need a different level of care after delivery.

It is not always possible to calculate the number of weeks accurately from the last menstrual period. Some people have irregular cycles or do not track their periods, so an early ultrasound can provide a more reliable reference.

The barrier it is trying to address

Around two-thirds of the world’s population does not have easy access to diagnostic imaging services, such as ultrasounds and X-rays. Traditional equipment is often large, expensive and difficult to repair in remote clinics. It also requires a stable power supply and specialized staff.

Portable devices have reduced some of these difficulties, but they do not solve the shortage of professionals who can use them. Google’s tested approach attempts to separate the two tasks: a local worker captures the images, and AI helps interpret them.

This is still research, not a general replacement for specialists or a guarantee that every complication can be detected with this method. But it shows one possible way to bring basic prenatal information to places where an ultrasound is currently out of reach for many families.

The next step will be to see how the tool performs in more communities, with different devices and medical situations. If it maintains its accuracy outside the study, AI could become a support layer for health workers caring for pregnancies in low-resource areas.