General Papers
Top studies that demonstrate the overall technology
💡 This is the key paper: trained on 200K fundus photos and validated on externally-tested 18K photos from clinics across China. The study was sponsored by Sun Yat-sen University and pre-registered at Clinicaltrials.gov.
The dataset was further validated on another 11K public domain images (Kaggle) from EyePACS made up of Hispanic (50%) as well as White, Black, and Asians.
The mean AUC for identifying the 14 retinal abnormalities and normal fundus was 0·955 (SD 0·046) in the internal validation set and 0·968 (0·037) in the external test set.
Retinal Age
Retinal age gap as a predictive biomarker for mortality risk
Zhu Z, Shi D, Guankai P, et al Retinal age gap as a predictive biomarker for mortality risk British Journal of Ophthalmology 2023; 107: 547-554.
Based on a study of 80K fundus images from the UK Biobank, 2006-2010. Results suggest overall mortality risk goes up 2% for each year increase in your retinal age. Up 3% for non-cardiovascular and non-cancer death.
Simon Nusinovici, Tyler Hyungtaek Rim, Marco Yu, Geunyoung Lee, Yih-Chung Tham, Ning Cheung, Crystal Chun Yuen Chong, Zhi Da Soh, Sahil Thakur, Chan Joo Lee, Charumathi Sabanayagam, Byoung Kwon Lee, Sungha Park, Sung Soo Kim, Hyeon Chang Kim, Tien-Yin Wong, Ching-Yu Cheng, Retinal photograph-based deep learning predicts biological age, and stratifies morbidity and mortality risk, Age and Ageing, Volume 51, Issue 4, April 2022, afac065, https://doi.org/10.1093/ageing/afac065
A team of Singapore researchers trained on 120K fundus photos from 40K Koreans and evaluated against 56K UK Biobank participants.