Singapore researchers have shown that advances artificial intelligence (AI) These techniques can significantly improve clinical diagnostics in resource-limited countries without the need for massive local datasets.
A team from Duke-NUS Medical School successfully applied transfer learninga method in which a model developed for one task is reused as a starting point for another, to predict patient outcomes after cardiac arrest.
The study, published in npj Digital Medicineaddresses a common challenge in AI adoption in low- and middle-income countries, namely the lack of comprehensive, high-quality data needed to train algorithmic models from scratch.
To test the effectiveness of transfer learning, the researchers used a brain recovery prediction model initially built in Japan using data from 46,918 out-of-hospital cardiac arrest patients. They adapted this model for use in Vietnam, testing it on a smaller group of 243 patients.
The results showed a huge improvement in diagnostic accuracy. When the original Japanese model was applied directly to the Vietnamese context, it distinguished high-risk from low-risk patients with an accuracy of 46%. However, the adapted transfer learning model achieved an accuracy rate of approximately 80%.
“The study shows that AI models do not need to be rebuilt from scratch for each new environment,” said Liu Nan, associate professor at the Center for Biomedical Data Science at Duke-NUS. “By adapting existing tools safely and efficiently, transfer learning can reduce costs, reduce development time, and help extend the benefits of AI to health systems with fewer resources.” »
Despite the growing potential of AI in healthcare, adoption of this technology remains uneven across the world. In a separate study published in Nature HealthDuke-NUS researchers and collaborators such as University College London (UCL) noted that while 63% of healthcare providers surveyed use AI tools, adoption is more widespread in high- and upper-middle-income countries.
Research has highlighted the potential for major language models (LLM) to improve access to care, diagnostics and clinical decision-making in low- and middle-income countries that continue to face barriers to adoption such as limited infrastructure and expertise.
An example is Sierra Leone, where community health workers use smartphone apps to detect malaria infections from blood smear samples, a more cost-effective method than conventional microscope-based systems. And in South Africa, chatbots provide prenatal counseling to pregnant women.
“LLMs have the greatest opportunity to transform healthcare in settings where specialist doctors are most scarce, but the global health community must urgently work together to ensure that the implementation of LLMs is supported in regions where adoption is most challenging,” said Siegfried Wagner of the UCL Institute of Ophthalmology and Moorfields Eye Hospital NHS Foundation Trust.
Ning Yilin, senior researcher at the Center for Biomedical Data Science at Duke-NUS, added that empowering people should be prioritized when integrating LLMs into healthcare.
“Strengthening digital literacy and building confidence in using these tools will ensure that AI supports the workforce, rather than disrupting it. Tailored skills development pathways can help under-resourced workers adapt and thrive, enabling AI to enhance and add value to clinical and administrative roles,” she said.
Call for international governance
Although AI tools have the potential to improve healthcare delivery, governance frameworks are essential to safe and ethical implementation of technology. Today, medical technology regulations often do not take into account Risks specific to AIlike privacy issues, hallucinations of modelssecurity and the need to monitor new tools.
To address these issues, Duke-NUS-led researchers proposed forming an international consortium called the Partnership for Oversight, Leadership and Accountability in the Regulation of Generative Models of Intelligent Systems in Medicine (Polaris-GM).
The consortium aims to provide guidance for regulating new tools, monitoring their impact, establishing security safeguards and adapting them to resource-constrained contexts. Bringing together healthcare industry leaders, regulators, ethicists and patient groups from around the world, Polaris-GM will review existing research before working towards a global consensus on AI governance in healthcare.
Jasmine Ong, from the Duke-NUS AI and Medical Sciences Initiative and a senior clinical pharmacist at Singapore General Hospital, said: “With clear oversight and clearly defined guidelines, health systems can confidently harness the many strengths of AI to improve health outcomes while avoiding potential pitfalls. From policymakers to patient groups, all stakeholders have a crucial role to play in making this goal a reality.”