CoChat
Sign Up
Cited by 339Retinal Imaging and AnalysisRetinal Diseases and TreatmentsOphthalmology and Visual Health Research

A deep learning system for predicting time to progression of diabetic retinopathy

Open access
Dai, Ling, Sheng, Bin, Chen, Tingli, Wu, Qiang, Liu, Ruhan, Cai, Chun, Wu, Liang, Yang, Dawei +44 more·2024·Nature Medicine
DOI 10.1038/s41591-023-02702-zPMC PMC10878973
Open this paper in CoChat — free accountOpen PDF

Abstract

Diabetic retinopathy (DR) is the leading cause of preventable blindness worldwide. The risk of DR progression is highly variable among different individuals, making it difficult to predict risk and personalize screening intervals. We developed and validated a deep learning system (DeepDR Plus) to predict time to DR progression within 5 years solely from fundus images. First, we used 717,308 fundus images from 179,327 participants with diabetes to pretrain the system. Subsequently, we trained and validated the system with a multiethnic dataset comprising 118,868 images from 29,868 participants with diabetes. For predicting time to DR progression, the system achieved concordance indexes of 0.754–0.846 and integrated Brier scores of 0.153–0.241 for all times up to 5 years. Furthermore, we validated the system in real-world cohorts of participants with diabetes. The integration with clinical workflow could potentially extend the mean screening interval from 12 months to 31.97 months, and the percentage of participants recommended to be screened at 1–5 years was 30.62%, 20.00%, 19.63%, 11.85% and 17.89%, respectively, while delayed detection of progression to vision-threatening DR was 0.18%. Altogether, the DeepDR Plus system could predict individualized risk and time to DR progression over 5 years, potentially allowing personalized screening intervals.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41591-023-02702-z

Cites

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • A deep learning-based AI system developed by Dai et al. could predict the progression of diabetic retinopathy by analyzing subtle changes in retinal microvasculature not easily detectable through traditional screening methods.

    “A study carried out by Dai et al. demonstrated that a deep learning-based AI system could predict the progression of diabetic retinopathy by analyzing subtle changes in retinal microvasculature that are not easily detectable through traditional screening methods (92,93).”

    From Generative artificial intelligence in predictive analysis of diabetes and its complications: a narrative review (2025) · cites this paper as [92]

    The Discussion reports that DeepDR Plus used baseline retinal images to predict individualized time to diabetic retinopathy progression. The Results (Explainability analysis) identify retinal vascular geometry as a possible predictive pattern, but say only that the model might detect signals beyond it. Our copy does not establish that the system predicts progression by detecting subtle microvascular changes that traditional screening methods cannot readily detect. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • The AI model was able to predict the likelihood of disease progression, improve image collection quality, provide clinical reference, and facilitate diabetic retinopathy screening with an accuracy rate that rivaled expert ophthalmologists.

    “The AI model was able to predict the likelihood of disease progression, improve image collection quality, provide clinical reference, and facilitate diabetic retinopathy screening with an accuracy rate that rivaled expert ophthalmologists, offering the potential for earlier intervention.”

    From Generative artificial intelligence in predictive analysis of diabetes and its complications: a narrative review (2025) · cites this paper as [92]

    The Results ('DeepDR Plus predicts time to DR progression'; page unavailable) report that the model predicted progression risk, and the Discussion describes its potential to guide personalized screening and clinical management. The Methods ('Image quality control and grading procedure'; page unavailable) describe ophthalmologists grading images and the exclusion of ungradable images, not AI improving image-collection quality. The paper does not report an accuracy comparison with expert ophthalmologists, and the Discussion says the system had not been applied in clinical practice. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

  • Generative AI models can integrate retinal image analysis with additional patient data, such as glycemic control history, blood pressure, and lipid levels, to provide a more comprehensive risk assessment.

    “Moreover, generative AI models can integrate retinal image analysis with additional patient data, such as glycemic control history, blood pressure, and lipid levels, to provide a more comprehensive risk assessment.”

    From Generative artificial intelligence in predictive analysis of diabetes and its complications: a narrative review (2025) · cites this paper as [92]

    Methods (Model development and training; no page number available) describes a combined model that uses retinal-image predictions alongside baseline HbA1c, blood pressure, and lipid measurements to predict diabetic-retinopathy progression. It does not describe a generative AI model or use glycemic-control history; its inputs are baseline measurements. Results (DeepDR Plus predicts time to DR progression) also says the combined model performed similarly to the retinal-image model, rather than establishing a more comprehensive risk assessment. (LLM, conf=0.90)

    Experimental — accuracy not yet measured

  • By factoring in additional data points, AI models can offer personalized predictions regarding the likelihood of retinopathy progression for each patient.

    “By factoring in these additional data points, AI models can offer personalized predictions regarding the likelihood of retinopathy progression for each patient (92).”

    From Generative artificial intelligence in predictive analysis of diabetes and its complications: a narrative review (2025) · cites this paper as [92]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

The citation lists haven’t been fetched yet

References 59 · Cited by 339

Research works printed but not linked: 56

  1. [1]Cheung N Mitchell P Wong TY Diabetic retinopathy Lancet 2010 376 124 136 20580421
  2. [2]Tan T-E Wong TY Diabetic retinopathy: looking forward to 2030 Front Endocrinol. 2022 13 1077669 PMC9868457 36699020
  3. [3]Wong TY Cheung CMG Larsen M Sharma S Simó R Diabetic retinopathy Nat. Rev. Dis. Prim. 2016 2 16012 27159554
  4. [4]Jenkins AJ Biomarkers in diabetic retinopathy Rev. Diabet. Stud. 2015 12 159 195 26676667 PMC5397989
  5. [5]Stratton IM UKPDS 50: risk factors for incidence and progression of retinopathy in type II diabetes over 6 years from diagnosis Diabetologia 2001 44 156 163 11270671
  6. [6]Solomon SD Diabetic Retinopathy: a position statement by the American Diabetes Association Diabetes Care 2017 40 412 418 28223445 PMC5402875
  7. [7]Wong TY Guidelines on diabetic eye care: the international council of ophthalmology recommendations for screening, follow-up, referral, and treatment based on resource settings Ophthalmology 2018 125 1608 1622 29776671
  8. [8]Flaxel CJ Diabetic retinopathy preferred practice pattern Ophthalmology 2020 127 P66 P145 31757498
  9. [9]Wang LZ Availability and variability in guidelines on diabetic retinopathy screening in Asian countries Br. J. Ophthalmol. 2017 101 1352 1360 28292772
  10. [10]Modjtahedi BS Two-year incidence of retinal intervention in patients with minimal or no diabetic retinopathy on telemedicine screening JAMA Ophthalmol. 2019 137 445 448 30730544 PMC6459096
  11. [11]Gunasekeran DV Ting DSW Tan GSW Wong TY Artificial intelligence for diabetic retinopathy screening, prediction and management Curr. Opin. Ophthalmol. 2020 31 357 365 32740069
  12. [12]Fundus photographic risk factors for progression of diabetic retinopathy. ETDRS report number 12. Early Treatment Diabetic Retinopathy Study Research Group. Ophthalmology98, 823–833 (1991). 2062515
  13. [13]Lund SH Individualised risk assessment for diabetic retinopathy and optimisation of screening intervals: a scientific approach to reducing healthcare costs Br. J. Ophthalmol. 2016 100 683 687 26377413 PMC4853547
  14. [14]Broadbent DM Safety and cost-effectiveness of individualised screening for diabetic retinopathy: the ISDR open-label, equivalence RCT Diabetologia 2021 64 56 69 33146763 PMC7716929
  15. [15]Bora A Predicting the risk of developing diabetic retinopathy using deep learning Lancet Digit. Health 2021 3 e10 e19 33735063
  16. [17]Ting DSW Artificial intelligence and deep learning in ophthalmology Br. J. Ophthalmol. 2019 103 167 175 30361278 PMC6362807
  17. [18]Ting DSW Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes JAMA 2017 318 2211 2223 29234807 PMC5820739
  18. [19]Gulshan V Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs JAMA 2016 316 2402 2410 27898976
  19. [20]Dai L A deep learning system for detecting diabetic retinopathy across the disease spectrum Nat. Commun. 2021 12 3242 34050158 PMC8163820
  20. [21]Arcadu F Deep learning algorithm predicts diabetic retinopathy progression in individual patients NPJ Digit. Med. 2019 2 92 31552296 PMC6754451
  21. [22]Cai C Effectiveness of quality of care for patients with type 2 diabetes in China: findings from the Shanghai Integration Model (SIM) Front. Med. 2022 16 126 138 34705246
  22. [23]Wilkinson CP Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales Ophthalmology 2003 110 1677 1682 13129861
  23. [24]Chen S A seven-year study on an integrated hospital-community diabetes management program in Chinese patients with diabetes Prim. Care Diabetes 2018 12 231 237 29396207
  24. [27]Raman R Incidence and progression of diabetic retinopathy in urban India: Sankara nethralaya-diabetic retinopathy epidemiology and molecular genetics study (SN-DREAMS II), Report 1 Ophthalmic Epidemiol. 2017 24 294 302 28332894
  25. [28]Varadarajan AV Deep learning for predicting refractive error from retinal fundus images Invest. Ophthalmol. Vis. Sci. 2018 59 2861 2868 30025129
  26. [29]Yang D Assessment of parafoveal diabetic macular ischemia on optical coherence tomography angiography images to predict diabetic retinal disease progression and visual acuity deterioration JAMA Ophthalmol. 2023 141 641 649 37227703 PMC10214181
  27. [30]Sun Z OCT angiography metrics predict progression of diabetic retinopathy and development of diabetic macular edema: a prospective study Ophthalmology 2019 126 1675 1684 31358386
  28. [31]Cheung CY A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre Nat. Biomed. Eng. 2021 5 498 508 33046867
  29. [32]CHEUNG CYL A new method to measure peripheral retinal vascular caliber over an extended area Microcirculation 2010 17 495 503 21040115
  30. [33]Scanlon PH Screening intervals for diabetic retinopathy and implications for care Curr. Diab. Rep. 2017 17 96 28875458 PMC5585285
  31. [36]Downing J Bollyky J Schneider J Use of a connected glucose meter and certified diabetes educator coaching to decrease the likelihood of abnormal blood glucose excursions: the livongo for diabetes program J. Med Internet Res. 2017 19 e234 28698167 PMC5527250
  32. [37]Poplin R Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning Nat. Biomed. Eng. 2018 2 158 164 31015713
  33. [38]Cheung CY-L Retinal vascular geometry and 6 year incidence and progression of diabetic retinopathy Diabetologia 2017 60 1770 1781 28623387
  34. [39]Lu Y Evaluation of automatically quantified foveal avascular zone metrics for diagnosis of diabetic retinopathy using optical coherence tomography angiography Invest. Ophthalmol. Vis. Sci. 2018 59 2212 2221 29715365 PMC5958306
  35. [40]Sawides L Alterations to the foveal cone mosaic of diabetic patients Invest. Ophthalmol. Vis. Sci. 2017 58 3395 3403 28687853 PMC5501497
  36. [41]Cheung CY Ikram MK Klein R Wong TY The clinical implications of recent studies on the structure and function of the retinal microvasculature in diabetes Diabetologia 2015 58 871 885 25669631
  37. [42]Klein R Retinal vascular abnormalities in persons with type 1 diabetes: the Wisconsin Epidemiologic Study of Diabetic Retinopathy: XVIII Ophthalmology 2003 110 2118 2125 14597518
  38. [43]Klein R The relation of retinal vessel caliber to the incidence and progressionof diabetic retinopathy: Xix: The Wisconsin Epidemiologic Study of Diabetic Retinopathy Arch. Ophthalmol. 2004 122 76 83 14718299
  39. [44]Oshitari T The pathogenesis and therapeutic approaches of diabetic neuropathy in the retina Int. J. Mol. Sci. 2021 22 9050 34445756 PMC8396448
  40. [45]Traversi C Fractal analysis of fluoroangiographic patterns in anterior ischaemic optic neuropathy and optic neuritis: a pilot study Clin. Exp. Ophthalmol. 2008 36 323 328 18700918
  41. [46]Simó R Hernández C European Consortium for the Early Treatment of Diabetic Retinopathy (EUROCONDOR) Neurodegeneration in the diabetic eye: new insights and therapeutic perspectives Trends Endocrinol. Metab. 2014 25 23 33 24183659
  42. [47]Zafar S Sachdeva M Frankfort BJ Channa R Retinal neurodegeneration as an early manifestation of diabetic eye disease and potential neuroprotective therapies Curr. Diab. Rep. 2019 19 17 30806815 PMC7192364
  43. [48]Sohn EH Retinal neurodegeneration may precede microvascular changes characteristic of diabetic retinopathy in diabetes mellitus Proc. Natl Acad. Sci. USA 2016 113 E2655 E2664 27114552 PMC4868487
  44. [49]Wong TY Sabanayagam C The war on diabetic retinopathy: where are we now? Asia Pac J. Ophthalmol. 2019 8 448 456 PMC6903323 31789647
  45. [50]Liu H Economic evaluation of combined population-based screening for multiple blindness-causing eye diseases in China: a cost-effectiveness analysis Lancet Glob. Health 2023 11 e456 e465 36702141
  46. [51]Yang Z Tan T-E Shao Y Wong TY Li X Classification of diabetic retinopathy: past, present and future Front. Endocrinol. 2022 13 1079217 PMC9800497 36589807
  47. [53]Majithia S Cohort profile: The Singapore Epidemiology of Eye Diseases study (SEED) Int. J. Epidemiol. 2021 50 41 52 33393587
  48. [54]Nguyen HV Cost-effectiveness of a National Telemedicine Diabetic Retinopathy Screening Program in Singapore Ophthalmology 2016 123 2571 2580 27726962
  49. [56]ElSayed NA 2. Classification and diagnosis of diabetes: Standards of Care in Diabetes—2023 Diabetes Care 2023 46 S19 S40 36507649 PMC9810477
  50. [57]Kaniadakis G The κ-statistics approach to epidemiology Sci. Rep. 2020 10 19949 33203913 PMC7673996
  51. [58]He, K., Fan, H., Wu, Y., Xie, S. & Girshick, R. Momentum contrast for unsupervised visual representation learning. in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition 9729–9738 (2020).
  52. [59]Chen, X., Fan, H., Girshick, R. & He, K. Improved baselines with momentum contrastive learning. arXiv preprint arXiv:2003.04297 (2020). · arXiv
  53. [60]Vaswani, A. et al. Attention is all you need. in Advances in neural information processing systems 5998–6008 (2017).
  54. [61]Zhao, H., Jia, J. & Koltun, V. Exploring self-attention for image recognition. in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition 10076–10085 (2020).
  55. [62]Lundberg SM Explainable machine-learning predictions for the prevention of hypoxaemia during surgery Nat. Biomed. Eng. 2018 2 749 760 31001455 PMC6467492
  56. [63]Lundberg, S. M. & Lee, S. -I. A unified approach to interpreting model predictions. in Advances in Neural Information Processing Systems30 (2017).

Authors
Dai, L., Sheng, B., Chen, T., Wu, Q., Liu, R., Cai, C., Wu, L., Yang, D., Hamzah, H., Liu, Y., Wang, X., Guan, Z., Yu, S., Li, T., Tang, Z., Ran, A., Che, H., Chen, H., Zheng, Y., Shu, J., Huang, S., Wu, C., Lin, S., Liu, D., Li, J., Wang, Z., Meng, Z., Shen, J., Hou, X., Deng, C., Ruan, L., Lu, F., Chee, M., Quek, T. C., Srinivasan, R., Raman, R., Sun, X., Wang, Y. X., Wu, J., Jin, H., Dai, R., Shen, D., Yang, X., Guo, M., Zhang, C., Cheung, C. Y., Tan, G. S. W., Tham, Y.-C., Cheng, C.-Y., Li, H., Wong, T. Y., & Jia, W.
Affiliations
  • Shanghai Jiao Tong University (CN)
  • Chinese University of Hong Kong (HK)
  • Singapore National Eye Center (SG)
  • Hong Kong University of Science and Technology (HK)
  • State Key Laboratory of Ophthalmology (CN)
  • Chinese Academy of Medical Sciences & Peking Union Medical College (CN)
  • Tongji Hospital (CN)
  • Huazhong University of Science and Technology (CN)
  • Sankara Nethralaya (IN)
  • Beijing Tongren Hospital (CN)
  • ShanghaiTech University (CN)
  • National University of Singapore (SG)
Corresponding author
  • Sheng, Bin
  • Li, Huating
  • Wong, Tien Yin
  • Jia, Weiping
ORCID
Venue
Nature Medicine
Publisher
Springer Science and Business Media LLC
Published
2024-01-04
Volume
30
Issue
2
Pages
584-594
ISSN
  • 1078-8956 (ISSN-L)
  • 1546-170X
DOI
10.1038/s41591-023-02702-z
PMC
PMC10878973
OpenAlex
W4390588537
Cited by
339
References
59
Citations by year
2026: 112 · 2025: 148 · 2024: 79
FWCI
86.36
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Medicine (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Rare & Special e-Zone (The Hong Kong University of Science and Technology) (submittedVersion)
Funders
  • National Natural Science Foundation of China (8238810007, 81870598)
  • Chinese Academy of Engineering
  • Young Scientists Fund
  • National Key Research and Development Program of China (2022YFA1004804, 2022YFC2502800)
Topics
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Retinal Imaging and Analysis
  • Health Sciences › Medicine › Ophthalmology › Retinal Diseases and Treatments
  • Health Sciences › Medicine › Ophthalmology › Ophthalmology and Visual Health Research
Keywords
diabetic retinopathy, fundus images, deep learning, vision-threatening diabetic retinopathy, diabetes mellitus, retinopathy screening, risk prediction
MeSH terms
Deep Learning, Blindness, Diabetes Mellitus, Diabetic Retinopathy, Humans

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review