A groundbreaking artificial intelligence model developed by Chinese researchers could revolutionize eye care accessibility worldwide. Scientists from Tsinghua University and Shanghai Jiao Tong University have unveiled EyeFM – a vision-based AI system trained on 14.5 million ocular images and clinical texts from diverse global populations.
Published in Nature Medicine, the international study involved 44 ophthalmologists across six countries validating the system’s effectiveness in both primary clinics and specialist hospitals. The model’s unique “doctor feedback” feature allows real-time adaptation, making it particularly valuable for low-resource settings.
In clinical trials with 668 high-risk patients, EyeFM demonstrated remarkable performance. When used as a diagnostic aid, it boosted ophthalmologists’ accuracy to 92.2% compared to 75.4% without AI support. This represents a significant leap from previous AI tools that lacked multi-ethnic training data and real-world clinical validation.
Dr. Li Wei, lead researcher from Tsinghua University, explains: “Unlike earlier systems limited to specific data types, EyeFM processes complex clinical information like human experts. It’s designed to complement – not replace – medical professionals.”
The development team emphasizes this technology could help address global disparities in eye care access, particularly in developing regions. With diabetes-related retinal diseases affecting over 140 million people worldwide, such innovations may prove crucial in preventing vision loss.
