Could AI ultrasounds be the 'great equalizer' for maternal health?

For many women, ultrasound scans during pregnancy are a routine part of antenatal care, offering them a first glimpse of their baby and helping doctors monitor crucial aspects of the fetus’ development, its position, and other factors which can mean the difference between life and death for a mother and her child. In low-income countries, however, this lifesaving technology remains inaccessible to millions, contributing to preventable maternal and infant deaths. The World Health Organization recommends at least one ultrasound during pregnancy, but in many of these nations, about 50% of pregnant women never receive a single scan, exacerbating the ongoing maternal and infant mortality crisis. WHO estimates that every year, 300,000 women die during pregnancy or childbirth, while over 2 million babies die within the first month of life, and 2 million are stillborn. Increasing the number of women who receive scans, and then connecting them to treatment where needed, could reduce these numbers. While the cost of ultrasound machines has dropped significantly in recent years, the lack of trained sonographers remains a major obstacle, particularly in remote areas. In low-income countries, there are fewer than two radiologists per 1 million people, compared to one per 10,000 in wealthier nations. However, a pioneering AI-enabled device — BabyChecker — may offer the solution. Developed by the Dutch social enterprise Delft Imaging, BabyChecker combines a handheld ultrasound probe which can deliver point-of-care ultrasound, or POCUS, with a smartphone app. The system guides health workers through a quick scan, capturing key images that are then analyzed by artificial intelligence. Trained to detect critical information such as gestational age, fetal position, and placenta location, the BabyChecker app can help determine whether a pregnancy is high-risk, medium-risk, or low-risk. High-risk mothers can then be referred to the nearest hospital for a specialist scan, while medium-risk mothers are usually invited back for another BabyChecker scan. This AI-driven system allows nonspecialist health workers, such as midwives or community health workers, to perform scans without needing a sonographer. The AI also shortens scan time from 45 minutes to just two minutes. “With Babychecker, every referral to a sonographer is justified, thus reducing traveling costs for mothers and ensuring our limited resources are used efficiently,” Dr. Gregory Ganda, county minister for health in Kisumu County, Kenya, where BabyChecker is being piloted, told Devex. While other companies are developing similar products, BabyChecker is “the first and only one of its kind AI-enabled POCUS device developed to screen risky pregnancies in low and middle-income countries,” Akshay Rajagopal, BabyChecker project manager for Delft Imaging in Kenya, told Devex. BabyChecker is also being trialed in Ethiopia, Honduras, Zambia, Sierra Leone, Ghana, Guatemala, Malawi, and Tanzania. The ‘great equalizer’ The Gates Foundation has been a key backer of AI-powered ultrasound technology, supporting initiatives such as Kenya’s obstetrics point-of-care ultrasound, or O-POCUS, guidelines. Gates grantees have purchased BabyChecker devices as part of their work in Pakistan and another in Kenya. Ari Moskowitz, deputy director of medical devices and AI at the Gates Foundation, described BabyChecker as a “great equalizer.” “The beauty of AI ultrasound is that you can take something which was previously specialized, expensive and diagnostic and put it in the hands of non-specialists, enabling them to do critical things in a quick and agile fashion,” Moskowitz said during a phone interview. Furthermore, BabyChecker is specifically designed to be used in remote, low-resource settings since the device doesn’t require constant internet or power to function. It’s also easy to use, with scanners only needing to watch a three-minute training video before they are ready to go. “The resource efficiency, portability and intuitive user-centric design of BabyChecker has facilitated its integration into remote healthcare settings where accessibility to healthcare services is a challenge. “This disruptive technology has accelerated task-shifting towards frontline health workers, and empowered them to screen risky pregnancies more effectively,” Rajagopal said in an email. Parents also get the chance to see their baby on the BabyChecker smartphone app, which can generate an ultrasound image after the scan. Promising results in Kenya One of the most advanced pilots is taking place in Kisumu County, Kenya, and is showing promising results, Ganda told Devex. The pilot, which started last November, has already scanned more than 700 women, detecting 57 high-risk and 159 mild-risk cases. The PharmAccess Foundation, a Dutch NGO, is covering the cost, which includes one handheld ultrasound probe, a smartphone, and a subscription to BabyChecker. The full kit costs around $5,000, which covers unlimited scans and support from Delft Imaging for a year. By comparison, a traditional mid-range ultrasound machine costs between $25,0000 and $45,000, Delft Imaging’s Rajagopal said. In Kisumu County, nurse midwife Teresa Ogada uses BabyChecker to scan women at their local community health facility at least twice during their pregnancy. Previously, most of these women would not be scanned at all due to long travel times to the nearest hospital, the high cost of transportation, and the limited availability of traditional ultrasound machines. For many, the journey could take hours, with long wait times once they arrived. “Before BabyChecker, it was a struggle,” Ogada explained. “We often knew a woman needed an ultrasound, but economic challenges kept her from getting to a hospital.” Thanks to BabyChecker, Ogada and her team have been able to detect complications early and refer mothers to the appropriate care. One of the most notable cases was a late-term pregnancy where the BabyChecker scan identified that the baby was in breech position. Thanks to the timely scan, the mother was referred to her nearest hospital, where she safely delivered her baby. If the mother had tried to deliver at home or at her local facility, both could have died, the nurse-midwife explained. However, the technology has limitations. The device requires regular charging, which can be a problem when health facilities experience power outages, something which happens fairly frequently, Ogada said. It also can’t pick up a fetal heartbeat or detect early pregnancy. Dr. Ganda agreed that BabyChecker will be more useful if it can do more, including measuring amniotic fluid levels, which is important for a healthy pregnancy, and detecting neural tube defects, such as spina bifida, which can be identified on ultrasound by observing a bulge on the back of the neck. “We want to do more with BabyChecker in the future, but the technology needs to catch up,” he said. Delft Imaging is working hard to meet these demands, including developing an AI algorithm to detect multiple gestation, estimated fetal weight, and amniotic fluid abnormalities, which are expected to be available in the application within a year, Rajagopal told Devex. The NGO is also working on how to measure the fetal heartbeat. Currently, midwives use low-technology solutions, including hand-held dopplers or fetoscopes, to listen to the baby’s heart. Cost an issue While BabyChecker is cheaper than traditional scanning options, cost is still an obstacle, Ganda said. Currently, the Kenya pilot uses one BabyChecker for one ward, which covers eight health facilities, with the device rotating between the facilities, scanning up to 25 women per day. While Ganda said he would like to see BabyChecker and other AI-enabled ultrasounds in all health facilities in Kisumu and other counties, this would only be realistic if the AI costs come down. “In terms of usability, BabyChecker is fantastic, and it has been able to identify fetal misrepresentation and low-lying placenta, which can cause hemorrhage. The biggest challenge is the operational expenditure,” Ganda said. “The technology charges bundles for use of the AI. But those numbers would ideally be reached in large, busy hospitals. In small facilities with less clients, the developers need to consider a low number of scans,” he explained. Enya Séguin from Delft Imaging agreed that the cost needs to come down. This is something which funders such as Gates can help with by supporting research — such as health economics studies into impact, outcomes, and cost-benefit analyses — into the potential benefits to communities and countries from improved maternal and antenatal screening with devices such as BabyChecker, she explained. “For us, the next step is sustainable implementation and integration of BabyChecker into health systems. This means finding the right access pricing and finding ways to bring down the costs of handheld ultrasound suppliers and motivate suppliers to be open to AI software integrations,” Séguin told Devex. Another potential barrier is opposition from some sonographers and radiographers, who see BabyChecker, and similar devices, as a threat, Ganda said. However, while Delft Imaging will continue to enhance BabyChecker, the idea is not to do sonographers out of a job. “The best ultrasound will always be done by a specialist,” Ganda said, adding that AI will never replace medical experts. “AI uses fixed learning markers, but medicine is not fixed,” he explained. Moskowitz from the Gates Foundation agreed that AI-enabled ultrasound can only do so much. “Products like BabyChecker are about task sharing, not taking the diagnostic task out of the hospital. It’s an early screening designed to better utilise the time of specialists and get women to the right places,” he said. For women like Aphline Aoko, who had two BabyChecker scans during her pregnancy in Kisumu before giving birth to her second baby, June Erica, in June, the device provided reassurance and helped detect risks that might otherwise have gone unnoticed. “We can see many things we didn’t know about the baby,” Aoko told Devex.