Global Initiative Launches First Universal Medical AI Foundation Model with 100-Country Collaboration

A pioneering international research coalition comprising over 100 study groups across more than 65 countries has announced the launch of the Global RETFound initiative. This groundbreaking project aims to develop the world's first truly globally representative Artificial Intelligence (AI) foundation model in medicine, utilizing an unprecedented dataset of 100 million eye images. The initiative builds on prior work with RETFound, the initial AI model for retinal and systemic disease detection, which was developed with a smaller dataset of 1.6 million fundus photographs.

Led by prominent institutions such as the National University of Singapore's Yong Loo Lin School of Medicine, Moorfields NHS Foundation Trust, University College London, and the Chinese University of Hong Kong, the project seeks to address the limitations of existing AI models that often rely on narrow, geographically confined data. The comprehensive dataset encompasses diverse populations from regions including Africa, the Middle East, South America, Southeast Asia, the Western Pacific, and the Caucasus, aiming to enhance the model's global applicability.

As detailed in a publication in Nature Medicine, the project represents one of the most extensive and diverse medical AI collaborations to date. The goal is to create a model capable of diagnosing a variety of ophthalmic and systemic diseases, such as diabetic retinopathy, glaucoma, age-related macular degeneration, and cardiovascular disorders. The model will be freely accessible under a Creative Commons license, promoting open research and development.

A key innovation of this effort is its dual data-sharing framework, designed to respect differing regulatory environments and resource capabilities. Participating institutions can either fine-tune local AI models with only the model weights shared centrally or contribute de-identified data through a secure infrastructure if they lack local GPU resources. This flexible approach enables broad participation regardless of institutional capacity.

Experts like Dr. Yih Chung Tham emphasize that current AI foundation models often reflect narrow demographic and geographic data bases, which can exacerbate health inequalities. The Global RETFound project seeks to mitigate this by fostering inclusive, resource-aware collaboration that prioritizes data privacy.

The initiative aims to set new standards for fairness and generalizability in medical AI, with a focus on usability across diverse healthcare settings. Although initially focused on ophthalmology, the methodology is designed to be adapted across other medical fields, paving the way for more equitable and effective AI tools globally.

This effort exemplifies how international cooperation can accelerate advancements in medical AI, ensuring broad and ethical benefits for healthcare worldwide. The consortium invites additional researchers and institutions to join in expanding this collaborative framework for more inclusive medical innovations.

For further information, refer to the publication: Yih Chung Tham et al., "Building the world's first truly global medical foundation model," Nature Medicine, 2025. source: https://medicalxpress.com/news/2025-09-global-medical-ai-foundation-country.html