Imagine going to an eye checkup and having an AI sidekick help your doctor spot issues faster than a Marvel hero spots villains 🦸♂️👁️!
That’s now a reality thanks to EyeFM, a vision-based foundation model developed by scientists from the Chinese mainland (Tsinghua University & Shanghai Jiao Tong University) in collaboration with global partners.
Published in Nature Medicine, EyeFM was trained on 14.5 million eye images paired with clinical notes from diverse, multi-ethnic datasets. In an international trial, 44 ophthalmologists from the Chinese mainland, India, Malaysia, Denmark, Equatorial Guinea and the United States validated its power as a clinical co-pilot 🌍💡.
In a randomized, double-masked trial at a center in China involving 668 high-risk patients, doctors using EyeFM hit a 92.2% diagnostic accuracy—up from 75.4% with standard care alone 🚀. That’s a game-changer for early retinal disease detection!
EyeFM also offers a “doctor feedback” feature, letting clinicians fine-tune the AI on the spot. This quick loop makes the model ideal for both high-complexity hospitals and low-resource clinics, bringing quality eye care to more communities.
Previous AI tools often learned from only one data type and lacked real-world testing. EyeFM blends images, text and real-time feedback, paving the way for everyday AI-powered eye exams. Stay tuned as AI reshapes healthcare across borders! 🌟
Reference(s):
cgtn.com
