A foundation model for generalizable disease detection from retinal images

Zhou, Yukun, Chia, Mark A., Wagner, Siegfried K., Ayhan, Murat S., Williamson, Dominic J., Struyven, Robbert R., Liu, Timing, Xu, Moucheng, Lozano, Mateo G., Woodward-Court, Peter, Kihara, Yuka, Allen, Naomi, Gallacher, John E. J., Littlejohns, Thomas, Aslam, Tariq, Bishop, Paul, Black, Graeme, Sergouniotis, Panagiotis, Atan, Denize, Dick, Andrew D., Williams, Cathy, Barman, Sarah, Barrett, Jenny H., Mackie, Sarah, Braithwaite, Tasanee, Carare, Roxana O., Ennis, Sarah, Gibson, Jane, Lotery, Andrew J., Self, Jay, Chakravarthy, Usha, Hogg, Ruth E., Paterson, Euan, Woodside, Jayne, Peto, Tunde, Mckay, Gareth, Mcguinness, Bernadette, Foster, Paul J., Balaskas, Konstantinos, Khawaja, Anthony P., Pontikos, Nikolas, Rahi, Jugnoo S., Lascaratos, Gerassimos, Patel, Praveen J., Chan, Michelle, Chua, Sharon Y. L., Day, Alexander, Desai, Parul, Egan, Cathy, Fruttiger, Marcus, Garway-Heath, David F., Hardcastle, Alison, Khaw, Sir Peng T., Moore, Tony, Sivaprasad, Sobha, Strouthidis, Nicholas, Thomas, Dhanes, Tufail, Adnan, Viswanathan, Ananth C., Dhillon, Bal, Macgillivray, Tom, Sudlow, Cathie, Vitart, Veronique, Doney, Alexander, Trucco, Emanuele, Guggenheim, Jeremy ORCID: https://orcid.org/0000-0001-5164-340X, Morgan, James E. ORCID: https://orcid.org/0000-0002-8920-1065, Hammond, Chris J., Williams, Katie, Hysi, Pirro, Harding, Simon P., Zheng, Yalin, Luben, Robert, Luthert, Phil, Sun, Zihan, McKibbin, Martin, O'Sullivan, Eoin, Oram, Richard, Weedon, Mike, Owen, Chris G., Rudnicka, Alicja R., Sattar, Naveed, Steel, David, Stratton, Irene, Tapp, Robyn, Yates, Max M., Petzold, Axel, Madhusudhan, Savita, Altmann, Andre, Lee, Aaron Y., Topol, Eric J., Denniston, Alastair K., Alexander, Daniel C. and Keane, Pearse A. 2023. A foundation model for generalizable disease detection from retinal images. Nature 622 (7981) , pp. 156-163. 10.1038/s41586-023-06555-x

Preview

PDF - Published Version Available under License Creative Commons Attribution. Download (22MB) | Preview

Abstract

Medical artificial intelligence (AI) offers great potential for recognizing signs of health conditions in retinal images and expediting the diagnosis of eye diseases and systemic disorders1. However, the development of AI models requires substantial annotation and models are usually task-specific with limited generalizability to different clinical applications2. Here, we present RETFound, a foundation model for retinal images that learns generalizable representations from unlabelled retinal images and provides a basis for label-efficient model adaptation in several applications. Specifically, RETFound is trained on 1.6 million unlabelled retinal images by means of self-supervised learning and then adapted to disease detection tasks with explicit labels. We show that adapted RETFound consistently outperforms several comparison models in the diagnosis and prognosis of sight-threatening eye diseases, as well as incident prediction of complex systemic disorders such as heart failure and myocardial infarction with fewer labelled data. RETFound provides a generalizable solution to improve model performance and alleviate the annotation workload of experts to enable broad clinical AI applications from retinal imaging.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Optometry and Vision Sciences
Publisher:	Nature Publishing Group
ISSN:	0028-0836
Date of First Compliant Deposit:	7 March 2024
Date of Acceptance:	18 August 2023
Last Modified:	20 Mar 2024 09:31
URI:	https://orca.cardiff.ac.uk/id/eprint/166980

Actions (repository staff only)

Edit Item