AI Tools Improve Diagnostics and Patient Prognoses in Resource-Limited Healthcare
Researchers at Duke-NUS Medical School have adapted AI models to enhance diagnostics and patient outcome prediction in resource-limited settings. The goal is to improve access to care and decision-making accuracy.
Singapore, January 27, 2026 – Research from Duke-NUS Medical School and collaborators demonstrates how artificial intelligence (AI) can improve healthcare delivery in challenging environments. The team has successfully adapted an advanced AI model to predict neurological recovery after cardiac arrest for patients in resource-limited healthcare settings.
The study, published in npj Digital Medicine, employs "transfer learning," an advanced AI technique that adapts pre-trained models, built on large datasets, for use in new environments with limited local data. This method is particularly relevant for low- and middle-income countries where data collection is constrained. An AI model originally developed in Japan, based on data from 46,918 patients, was adapted for a cohort of 243 patients in Vietnam.
The adapted model significantly improved accuracy, distinguishing between high-risk and low-risk patients with approximately 80 percent accuracy, compared to the original model's approximately 46 percent accuracy in the Vietnamese context. Associate Professor Liu Nan from Duke-NUS highlighted that transfer learning can reduce costs, shorten development time, and extend AI benefits to healthcare systems with fewer resources.
AI's potential extends to other applications. In a separate study published in Nature Health, researchers from Duke-NUS and University College London (UCL) examined how large language models (LLMs) could advance global health. LLMs have the potential to enhance access to care, diagnostics, and clinical decision-making in regions with specialist physician shortages. Examples include a pregnancy information chatbot and mobile applications for malaria detection.
However, the researchers noted that AI development remains concentrated in high-income countries. Many low- and middle-income countries continue to face barriers such as insufficient infrastructure and expertise. Additionally, clearer regulatory frameworks are needed to ensure the safe and ethical implementation of AI. A research group led by Duke-NUS is proposing the establishment of an international consortium to guide AI regulation in healthcare.