RETFound is published by Nature! Congrats to the leading authors Yukun and Pearse!
“A foundation model for generalizable disease detection from retinal images” has been accepted and published at Nature (Impact factor: 64) Special congrats to the first-author Yukun Zhou who did a truly amazing job!
In the paper, we present the first generative modelling based foundational model in medicine. The proposed RETFound is based on masked auto-encoder, which was pretrained on natural images first, followed by further pre-training on 1.6 milliion retinal images. RETFound shows great generaisation on more than 10 down stream tasks after fine tuning on very limited amount of task specific data. ***