Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2022 Mar;2(3):264-271.
doi: 10.1038/s43587-022-00171-6. Epub 2022 Feb 21.

Detecting visually significant cataract using retinal photograph-based deep learning

Yih-Chung Tham #  1   2   3 , Jocelyn Hui Lin Goh #  1 , Ayesha Anees #  4 , Xiaofeng Lei #  4 , Tyler Hyungtaek Rim  1   2 , Miao-Li Chee  1 , Ya Xing Wang  5 , Jost B Jonas  6 , Sahil Thakur  1 , Zhen Ling Teo  1 , Ning Cheung  1   2 , Haslina Hamzah  1 , Gavin S W Tan  1   2 , Rahat Husain  1   2 , Charumathi Sabanayagam  1   2 , Jie Jin Wang  2 , Qingyu Chen  7 , Zhiyong Lu  7 , Tiarnan D Keenan  8 , Emily Y Chew  8 , Ava Grace Tan  9   10 , Paul Mitchell  9 , Rick S M Goh  4 , Xinxing Xu #  4 , Yong Liu #  1   2   4 , Tien Yin Wong #  1   2   3 , Ching-Yu Cheng #  11   12   13
Affiliations

Detecting visually significant cataract using retinal photograph-based deep learning

Yih-Chung Tham et al. Nat Aging. 2022 Mar.

Erratum in

Abstract

Age-related cataracts are the leading cause of visual impairment among older adults. Many significant cases remain undiagnosed or neglected in communities, due to limited availability or accessibility to cataract screening. In the present study, we report the development and validation of a retinal photograph-based, deep-learning algorithm for automated detection of visually significant cataracts, using more than 25,000 images from population-based studies. In the internal test set, the area under the receiver operating characteristic curve (AUROC) was 96.6%. External testing performed across three studies showed AUROCs of 91.6-96.5%. In a separate test set of 186 eyes, we further compared the algorithm's performance with 4 ophthalmologists' evaluations. The algorithm performed comparably, if not being slightly more superior (sensitivity of 93.3% versus 51.7-96.6% by ophthalmologists and specificity of 99.0% versus 90.7-97.9% by ophthalmologists). Our findings show the potential of a retinal photograph-based screening tool for visually significant cataracts among older adults, providing more appropriate referrals to tertiary eye centers.

PubMed Disclaimer

Conflict of interest statement

T.H.R. was a scientific advisor to a start-up company called Medi Whale; he received stock as a part of the standard compensation package. T.H.R. also reports personal fees from Allergan and Novartis, and patents pending for: cardiovascular disease diagnosis assistant method and apparatus (10–2018–0166720(KR), 10–2018–0166721(KR), 10–2018–0166722(KR) and PCT/KR2018/016388); diagnosis assistance system (10–2018–0157559(KR), 10–2018–0157560(KR) and 10–2018–0157561(KR)); diagnosis technology using AI (62/694,901 (USA) and 62/776,345 (USA)); method for controlling a portable fundus camera and diagnosing disease using the portable fundus camera (62/715,729 (USA)); and method for predicting cardio-cerebrovascular disease using eye image (10–2017–0175865 (K.R.)). T.Y.W. is a consultant and a member of the advisory boards for Allergan, Bayer, Boehringer Ingelheim, Genentech, Merck, Novartis, Oxurion (formerly ThromboGenics), Roche and Samsung Bioepis, and cofounder of the start-up companies Plano Pte and EyRiS. T.Y.W. also has a patent issued for Deep Learning System for Retinal Diseases (PCT/SG2018/050363, Singapore and 10201901218S (provisional) Singapore). All the other authors declare no competing interests.

Figures

Fig. 1
Fig. 1
ROC curve showing performance of the classification algorithm for the detection of visually significant cataracts (defined as BCVA < 20/60).
Fig. 2
Fig. 2. Saliency maps highlighting regions that the algorithm focuses on when predicting visually significant cataracts.
The highlighted regions in retinal photographs are congruent with the pathological features that typically present in eyes with significant cataracts. Cataract eyes with localized haze (a) and generalized haze (b) presented on retinal photos.
Fig. 3
Fig. 3
ROC curve showing performance of the algorithm versus 2 professional graders and 4 ophthalmologists on a test set of 186 eyes (randomly selected from SCES and SINDI).
See this image and copyright information in PMC

References

    1. Adelson, J. D. et al. Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the Right to Sight: an analysis for the Global Burden of Disease Study. Lancet Global Health10.1016/S2214-109X(20)30489-7 (2020). - DOI - PMC - PubMed
    1. Chua J, et al. Prevalence, risk factors, and impact of undiagnosed visually significant cataract: the Singapore epidemiology of eye diseases study. PLoS ONE. 2017;12:e0170804. doi: 10.1371/journal.pone.0170804. - DOI - PMC - PubMed
    1. Keel S, McGuiness MB, Foreman J, Taylor HR, Dirani M. The prevalence of visually significant cataract in the Australian National Eye Health Survey. Eye. 2019;33:957–964. doi: 10.1038/s41433-019-0354-x. - DOI - PMC - PubMed
    1. Lansingh VC, Carter MJ, Martens M. Global cost-effectiveness of cataract surgery. Ophthalmology. 2007;114:1670–1678. doi: 10.1016/j.ophtha.2006.12.013. - DOI - PubMed
    1. Shrime MG, et al. Cost-effectiveness in global surgery: pearls, pitfalls, and a checklist. World J. Surg. 2017;41:1401–1413. doi: 10.1007/s00268-017-3875-0. - DOI - PubMed

Publication types