Watch the recording.
On September 9, 2025, the Alliance hosted AI-Assisted Ultrasound: Bridging Innovation and Access in Low-Resource Settings as part of its 2025 AI and Global Health Discussion Series, sponsored by Pfizer. The panel brought together perspectives from across sectors — academics, technology innovators, clinical implementers, funders, and clinicians — including representatives from AMPATH Kenya, the Gates Foundation, the Global Ultrasound Institute (GUSI), Philips Ultrasound, and UCSF’s Bixby Center for Global Reproductive Health.
Together, the expert panelists explored how AI-assisted ultrasound can transform care in low-resource settings while addressing the challenges of implementation. They discussed its potential to democratize diagnostics — making affordable imaging more accessible for maternal health, cancer, cardiovascular disease, tuberculosis, and other critical conditions where access is too often limited.
Dr. Dilys Walker, director of global research at UCSF’s Bixby Center, framed the discussion by underscoring both the transformative potential and the important questions to consider surrounding AI-assisted ultrasound. “With AI-assisted point-of-care ultrasound (POCUS), we have a powerful step forward with the potential to transform access and ultimately save lives,” she said.
At the same time, Dr. Walker urged the audience to look beyond just the promise and consider the implications of how the technology is developed and deployed. “[AI POCUS] raises serious questions: Are we choosing the right use cases for algorithm development? Who gets to decide which problems AI POCUS should solve? And are designers truly working with limited-resource settings in mind — or might we unintentionally be widening inequities?”
She reminded participants that this is both a moment of optimism and caution. “We are at a truly exciting — but also challenging — moment. The promise of AI POCUS is real, but so are the risks.”
“We are at a truly exciting — but also challenging — moment. The promise of AI POCUS is real, but so are the risks.” – Dr. Dilys Walker, UCSF’s Bixby Center
Key Takeaways:
- AI POCUS should aim to bring diagnostics to patients, rather than the other way around.
- Affordable, multipurpose AI devices can detect risk and encourage care-seeking behavior.
- AI POCUS should empower clinicians, not replace them.
- Local context and health system integration are critical.
- AI’s impact depends on multi-use design and diverse, locally relevant data.
- Trust and pacing are essential for meaningful adoption.
The Funder Perspective: Scaling Access and Reducing Costs
Richard Zong, program officer – medical device & AI development at the Gates Foundation, highlighted why the foundation has made AI-assisted ultrasound a priority investment.
“Training takes time and effort, and point-of-care ultrasound is relatively expensive for what these [low-resource] settings can afford…we wanted to lower the cost of ultrasound and increase accessibility, such that in the long run, people everywhere will have access to the power of ultrasound,” Zong explained, highlighting how in many low- and middle-income countries, particularly at the primary care level, ultrasound remains out of reach.
Zong shared how the Gates Foundation sees value not only in AI’s capacity to detect risk, but also in its potential to strengthen care-seeking behavior. “We’re really excited about this AI-assisted ultrasound tool not only for its ability to detect risk, but also its ability to bring people to the clinic,” he said, describing a “magnet effect” in which greater accessibility to ultrasound encourages people to come in more frequently and receive more treatment and care.
The ultimate aim, he added, is to maximize both value and reach by building multipurpose devices that integrate several AI applications. “Our goal is to continue partnering with manufacturers to innovate on these solutions so they become lower and lower in cost. As I look towards the future, I’m incredibly excited about how much innovation we can put into AI-assisted ultrasound platforms,” he said, noting the potential and interest to expand beyond obstetrics into lung screening for tuberculosis and pneumonia, gynecological conditions, cardiac care, and even breast cancer.
Implementation and Building Capacity at Scale
“Democratizing and decentralizing of imaging allows the clinician to make an earlier, less delayed, less costly, and more upstream diagnosis for that patient … and AI is definitely going to accelerate that and make POCUS even more accessible.” – Dr. Kevin Bergman, GUSI
Dr. Kevin Bergman, co-founder and co-CEO of the Global Ultrasound Institute (GUSI), which has led education and training for the three largest POCUS projects globally, shared his perspective on the current opportunities and biggest risks associated with AI POCUS.
“The fundamental transformation of POCUS is that it brings the diagnostic imaging to where the patient is, instead of sending the patient to where the diagnostic imaging is,” shared Dr. Bergman. “Democratizing and decentralizing of imaging allows the clinician to make an earlier, less delayed, less costly, and more upstream diagnosis for that patient … and AI is definitely going to accelerate that and make POCUS even more accessible.”
Dr. Bergman emphasized five important considerations for integrating AI responsibly into POCUS:
- AI should be framed and built as an assist for human clinicians and not a replacement for them.
- AI point-of-care ultrasound needs to be designed to build trust. To achieve adoption and ultimately impact, it’s paramount that clinicians trust this tool.
- AI is an accelerator of learning ultrasound, not a replacement for knowledge or skills.
- Use diverse data sets with local feedback loops.
- Findings on a diagnostic imaging report do not equal a diagnosis; in POCUS, context is everything.
“If we build AI for POCUS that empowers, educates, and includes clinicians, we’ll create stronger, more trusted healthcare,” Dr. Bergman added. “Done right, AI POCUS won’t replace clinicians; they’ll make them faster, better, and more capable.”
“If we build AI for POCUS that empowers, educates, and includes clinicians, we’ll create stronger, more trusted healthcare. Done right, AI POCUS won’t replace clinicians; they’ll make them faster, better, and more capable.” – Dr. Kevin Bergman, GUSI
Initiating AI-Enabled POCUS Rollout: AMPATH’s Lessons in Western Kenya
AMPATH Kenya’s Dr. Hussein Elias, family physician and lecturer in the Department of Family Medicine at Moi University, and Dr. Daria Szkwarko, family physician and associate professor in the Department of Family Medicine at the Warren Alpert Medical School of Brown University, co-lead one of the largest multi-use point-of-care ultrasound programs in the world, with the goal of deploying more than 600 ultrasound probes and training over 3,000 nurses, midwives, clinical officers, and medical officers across six counties in western Kenya.
Close collaboration with the Ministry of Health has been key in AMPATH’s work, including county-level agreements detailing all of the steps of focus, education, and clinical integration. The program spans a range of use case topics, including cardiopulmonary disease (particularly heart failure), second- and third-trimester obstetric complications, and breast cancer assessment for patients presenting with palpable masses. “We selected the topics for multiple use cases by considering morbidity and mortality in the rural region in Western Kenya where we practice,” Dr. Szkwarko shared.
“Healthcare workers and patients must trust the technology being integrated — and trust takes time, thoughtful communication, transparency, and reliability.” – Dr. Daria Szkwarko, AMPATH Kenya, Brown University
The team has encountered two main challenges while implementing multi-use POCUS without AI, with plans to integrate AI in the coming months.
“Healthcare workers are incredibly busy … taking the extra time to document findings in patient books is very challenging. Referral outcomes are also difficult to track at baseline … without investing in additional resources,” explained Dr. Szkwarko, sharing how while they have prioritized sustainability with country-led documentation, this approach has often led to documentation gaps between scans that are performed and those that are actually recorded.
The team also cautioned that even with expanded diagnostics, broader systemic barriers persist. “Although we are implementing an incredible diagnostic tool, we are not changing the state of other social determinants of health in rural Kenya or the state of the health system,” Dr. Elias said, pointing to barriers such as patients’ inability to afford travel or limited availability of biopsies at higher-level facilities.
As they prepare to integrate AI-assisted ultrasound in the coming months, both emphasized the importance of pacing and trust. “We’re only as good as our health system infrastructure, and we can only move at the speed of trust,” Dr. Szkwarko stressed. “Healthcare workers and patients must trust the technology being integrated — and trust takes time, thoughtful communication, transparency, and reliability.”
“Although we are implementing an incredible diagnostic tool, we are not changing the state of other social determinants of health in rural Kenya or the state of the health system.” – Dr. Hussein Elias, AMPATH Kenya, Moi University
Product Development with Patients and Providers in Mind
Jonathan Sutton, senior scientist and group lead at Philips Ultrasound, shared his industry perspective on developing an AI-assisted ultrasound product. “We perceive that there is a major gap in care providers who can consistently assess high-risk pregnancies, and closing it is difficult due to training costs, limited personnel, and uneven health system infrastructure,” Sutton said.
In collaboration with the Gates Foundation, Philips is building a product around that perceived clinical unmet need. “We haven’t released anything to the market yet, but this SmartSweep workflow builds on our handheld Lumify ultrasound platform. After a few hours of training, a clinical user — whether a nurse, midwife, clinical officer, or even a sonographer or OB-GYN — can perform an abdominal scan guided by onscreen prompts and live feedback to ensure the user is following proper ultrasound practices,” shared Sutton.
One thing they are emphasizing is ensuring quality when machine learning is actually involved, with good machine learning practices becoming an essential tenet of their product development process.
“In product development, we focus on understanding user needs by interacting and iterating with clinicians and stakeholders who are actually in the clinical environments. These insights shape our system requirements and software and hardware specifications. We then iteratively test the product with market research, including usability, system, and AI performance validation, before release. We intend to create a platform that can continue improving in accuracy, features, and overall fit for users,” explained Sutton.
Collaboration and Trust: Key to AI-Assisted Ultrasound in Low-Resource Settings
“It’s important to think about the contextually relevant AI solutions that can be integrated into the existing infrastructure, which can then be monitored, evaluated, and regulated. Every country is unique in its own way, and it’s about identifying what is truly relevant for each country and how we can put it together so we don’t create parallel systems, but instead work within existing systems to strengthen them.” – Dr. Hussein Elias, AMPATH Kenya, Moi University
While AI-assisted ultrasound technology has the potential to expand access to diagnostics in low-resource settings, its success will hinge on affordability, adequate training, seamless clinical integration, and strengthened referral pathways. A consistent theme emerged across the panel: collaboration and trust are essential to harnessing AI effectively in healthcare.
“It’s important to think about the contextually relevant AI solutions that can be integrated into the existing infrastructure, which can then be monitored, evaluated, and regulated. Every country is unique in its own way, and it’s about identifying what is truly relevant for each country and how we can put it together so we don’t create parallel systems, but instead work within existing systems to strengthen them,” emphasized Dr. Elias.
Panelists also cautioned that AI and machine learning, though promising, are not silver bullets. Stakeholders must remain both optimistic and critical, championing responsible use and ongoing education. Moving deliberately, leveraging existing systems, and prioritizing contextually relevant solutions are crucial to strengthening healthcare infrastructure. Trust, which is built through people, partnerships, and keeping patients at the center, remains key to meaningful adoption.
Watch the recording.
Check out the previous convenings in our 2024 AI and Global Health Discussion Series, and our other work shaping AI and global health here. With thanks to Pfizer for sponsoring the 2025 AI and Global Health Discussion Series, and to Samsung Research America for its support of this AI-Assisted Ultrasound convening.
Bios
Kevin Bergman, Co-Founder and Co-CEO, Global Ultrasound Institute (GUSI)
Dr. Kevin Bergman is the co-founder and co-CEO of Global Ultrasound Institute, a physician, educator, and entrepreneur. He is a strong advocate for adoption of point of care ultrasound across the continuum of care, with a particular focus on primary care, global health, and health equity.
Dr. Bergman is the immediate past co-Director of the Ultrasound and Global Health programs at the UCSF Contra Costa Family Medicine Residency, a faculty member in the family medicine residency program, and previously served as an attending physician in the emergency department at Contra Costa County Hospital from 2009 to 2022. Dr. Bergman is a member of the Board of Directors of the Society of Ultrasound in Medical Education and an official consultant to the International Consensus Conference on Ultrasound in Medical Education. He was the founding Vice-Chair of the American Academy of Family Physicians (AAFP) Ultrasound Member Interest Group and co-authored the AAFP POCUS guidelines and an Associate Clinical Professor in the UCSF Department of Family and Community Medicine. Since 2011, Dr. Bergman has taught POCUS at dozens of national and international conferences, including WINFOCUS, AIUM/SUSME, AAFP, EM Essentials, and the Society of Teachers of Family Medicine.
Hussein Elias, Family Physician and Lecturer in the Moi University Department of Family Medicine, Medical Education and Community Health, AMPATH Kenya
Dr. Elias is a Family Medicine consultant, currently holding a position of a faculty member at the Department of Family Medicine, Medical Education and Community Health, Moi University. He is engaged in care, education and research and has been working with AMPATH for the past several years. During his training in Family Medicine, most of his direct clinical experience occurred in the context of the primary care networks. This is the context in which he continues to provide direct clinical care and where he has developed new systems to enhance continuity of care. In collaboration with Dr. Daria Szkwarko and with the support of AMPATH, he currently serves as the co-Principal Investigator for Implementing Point of Care Ultrasound (POCUS) Education and Care Integration throughout Western Kenya to Enhance Patient-Centered Primary Care, a multi-year and innovative POCUS implementation for primary care facilities in western Kenya focusing on breast, obstetrics and cardiopulmonary.
Jonathan Sutton, Senior Scientist & Group Lead, Philips
Jonathan Sutton is a Senior Scientist and Group Leader at Philips Ultrasound in Cambridge, USA, and Clinical Science Lead for Philips’ AI-Enabled Obstetric Application Suite initiative with the Gates Foundation.
His work in the early innovation arm of Philips has helped bring products to market across several domains including cerebral and cardiovascular ultrasound imaging, patient monitoring, and pre-hospital care delivery. Jonathan has leveraged his technical background to lead a variety of programs within Philips ranging from core machine learning and imaging interoperability technology development to value proposition field research. Prior to joining Philips, Jonathan has a background in physics and economics from the College of William and Mary, earned a Ph.D. in Biomedical Engineering from the University of Cincinnati, and performed postdoctoral fellowship work at Harvard Medical School. Jonathan’s expertise lies at the intersection of ultrasound physics and human anatomy and physiology.
Daria Szkwarko, Associate Professor, Department of Family Medicine at the Warren Alpert Medical School of Brown University, AMPATH Kenya
Daria Szkwarko, DO, MPH is an Associate Professor (teaching scholar track) in the Department of Family Medicine and the Director of the Global Health Faculty Development Fellowship in Family Medicine at the Warren Alpert Medical School at Brown University. Her research interests are in implementing and evaluating educational innovations that help democratize knowledge and improve healthcare access for vulnerable populations globally. Currently, Dr. Szkwarko co-leads the largest multiple use case point of care ultrasound (POCUS) project in the world which launched in August 2024. With funding provided through Panorama Global, with funding from Gates Foundation, Novartis, Eli Lilly and Company, and MSD through its MSD for Mothers initiative, Dr. Szkwarko and Dr. Hussein Elias (Moi University Family Medicine faculty) are aiming to train more than 3000 healthcare workers across western Kenya in POCUS. Dr. Szkwarko is a global health expert and is the Family Medicine Lead for the Academic Model Providing Access to Healthcare (AMPATH), one of the largest academic global health collaborations. At the Warren Alpert Medical School, Dr. Szkwarko serves as the Co-Director for the Reproductive Health Scholarly Concentration and the Co-Director of the Health Systems Science 2 and 4 courses for the Primary Care and Population Medicine program. Clinically, she is a hospitalist at Kent Hospital in Warwick, RI and does urgent care and TB infection management at Blackstone Valley Community Health Care in Central Falls, RI.
Dilys Walker, Professor, OB/GYN, Reproductive Sciences and Director of Global Health Research, UCSF Bixby Center for Global Reproductive Health
Dilys Walker is a Professor in the Department of Obstetrics, Gynecology and Reproductive Sciences at UCSF Institute for Global Health Sciences (IGHS). She serves as the Director of Global Health Research for the Bixby Center for Global Reproductive Health and the Director of the Institute for Global Health Science Center for Global Maternal Newborn Child Health Research. As a clinician researcher, her academic career has focused on implementation research in limited resource settings, taking a life course approach to improving outcomes for women, mothers, and their newborns through pregnancy to childbirth. Dr. Walker led efforts to evaluate a group model of antenatal care in Rwanda and another study in Kenya and Uganda to assess the effectiveness of an intervention package to improve quality of intrapartum care and decrease preterm neonatal mortality. Currently she is working with Dr Grace Githemo and her team at Kenyatta University to evaluate the implementation of POCUS in 8 Kenyan counties.
Richard Zong, Program Officer – Medical Device & AI Development, Gates Foundation
Richard is Program Officer – Medical Device & AI Development at the Gates Foundation and sits on the Devices and AI team that focuses on the introduction of maternal and newborn health innovations, with the aim of reducing mortality and morbidity in low- and middle-income countries through pregnancy risk stratification. The team invests in areas including AI-assisted ultrasound imaging and wearable sensing platforms to enable appropriate planning and referrals through the pregnancy cycle.