这是indexloc提供的服务，不要输入任何密码
Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Detection of signs of disease in external photographs of the eyes via deep learning

Nature Biomedical Engineering volume 6, pages 1370–1383 (2022)Cite this article

Subjects

Abstract

Retinal fundus photographs can be used to detect a range of retinal conditions. Here we show that deep-learning models trained instead on external photographs of the eyes can be used to detect diabetic retinopathy (DR), diabetic macular oedema and poor blood glucose control. We developed the models using eye photographs from 145,832 patients with diabetes from 301 DR screening sites and evaluated the models on four tasks and four validation datasets with a total of 48,644 patients from 198 additional screening sites. For all four tasks, the predictive performance of the deep-learning models was significantly higher than the performance of logistic regression models using self-reported demographic and medical history data, and the predictions generalized to patients with dilated pupils, to patients from a different DR screening programme and to a general eye care programme that included diabetics and non-diabetics. We also explored the use of the deep-learning models for the detection of elevated lipid levels. The utility of external eye photographs for the diagnosis and management of diseases should be further validated with images from different cameras and patient populations.

Access through your institution
Buy or subscribe

This is a preview of subscription content, access via your institution

Access options

Access through your institution

Buy this article

  • Purchase on SpringerLink
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Extracting insights from external photographs of the front of the eye.
Fig. 2: Importance of different regions of the image and impact of image resolution.
Fig. 3: Qualitative and quantitative saliency analysis illustrating the influence of various regions of the image towards the prediction.

Similar content being viewed by others

Data availability

The main data supporting the results in this study are available within the paper and its Supplementary Information. This study used de-identified data from EyePACS Inc. and the teleretinal diabetes screening programme at the Atlanta Veterans Affairs. Interested researchers should contact J.C. (jcuadros@eyepacs.com) to enquire about access to EyePACS data and approach the Office of Research and Development at https://www.research.va.gov/resources/ORD_Admin/ord_contacts.cfm to enquire about access to VA data.

Code availability

The deep-learning framework (TensorFlow) used in this study is available at https://www.tensorflow.org; the neural network architecture is available from https://github.com/tensorflow/models/blob/master/research/slim/nets/inception_v3.py; and an ImageNet pretrained checkpoint is available from https://github.com/tensorflow/models/tree/master/research/slim.

References

  1. Poplin, R. et al. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nat. Biomed. Eng. 2, 158–164 (2018).

    Article  Google Scholar 

  2. Cheung, C. Y. et al. A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre. Nat. Biomed. Eng. https://doi.org/10.1038/s41551-020-00626-4 (2020).

  3. Mitani, A. et al. Detection of anaemia from retinal fundus images via deep learning. Nat. Biomed. Eng. 4, 18–27 (2020).

    Article  Google Scholar 

  4. Sabanayagam, C. et al. A deep-learning algorithm to detect chronic kidney disease from retinal photographs in community-based populations. Lancet Digit. Health 2, e295–e302 (2020).

    Article  Google Scholar 

  5. Rim, T. H. et al. Prediction of systemic biomarkers from retinal photographs: development and validation of deep-learning algorithms. Lancet Digit. Health 2, e526–e536 (2020).

    Article  Google Scholar 

  6. Tarlan, B. & Kiratli, H. Subconjunctival hemorrhage: risk factors and potential indicators. Clin. Ophthalmol. 7, 1163–1170 (2013).

    Google Scholar 

  7. Hreidarsson, A. B. Pupil size in insulin-dependent diabetes. Relationship to duration, metabolic control, and long-term manifestations. Diabetes 31, 442–448 (1982).

    Article  CAS  Google Scholar 

  8. Smith, S. E., Smith, S. A., Brown, P. M., Fox, C. & Sonksen, P. H. Pupillary signs in diabetic autonomic neuropathy. Br. Med. J. 2, 924–927 (1978).

    Article  CAS  Google Scholar 

  9. Banaee, T. et al. Distribution of different-sized ocular surface vessels in diabetics and normal individuals. J. Ophthalmic Vis. Res. 12, 361–367 (2017).

    Article  Google Scholar 

  10. Iroshan, K. A. et al. Detection of diabetes by macrovascular tortuosity of superior bulbar conjunctiva. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. 2018, 1–4 (2018).

    CAS  Google Scholar 

  11. Comprehensive Diabetes Care. National Committee for Quality Assurance https://www.ncqa.org/hedis/measures/comprehensive-diabetes-care/ (2020).

  12. Zangemeister, W. H., Gronow, T. & Grzyska, U. Pupillary responses to single and sinusoidal light stimuli in diabetic patients. Neurol. Int. 1, e19 (2009).

    Article  Google Scholar 

  13. Hashemi, H. et al. White-to-white corneal diameter distribution in an adult population. J. Curr. Ophthalmol. 27, 21–24 (2015).

    Article  Google Scholar 

  14. Rüfer, F., Schröder, A. & Erb, C. White-to-white corneal diameter: normal values in healthy humans obtained with the Orbscan II topography system. Cornea 24, 259–261 (2005).

    Article  Google Scholar 

  15. Worthen, D. M., Fenton, B. M., Rosen, P. & Zweifach, B. Morphometry of diabetic conjunctival blood vessels. Ophthalmology 88, 655–657 (1981).

    Article  CAS  Google Scholar 

  16. Danilova, A. I. Blood circulation in the conjunctival blood vessels of patients with diabetes mellitus. Probl. Endokrinol. 26, 9–14 (1980).

    CAS  Google Scholar 

  17. Fenton, B. M., Zweifach, B. W. & Worthen, D. M. Quantitative morphometry of conjunctival microcirculation in diabetes mellitus. Microvasc. Res. 18, 153–166 (1979).

    Article  CAS  Google Scholar 

  18. Owen, C. G. et al. Diabetes and the tortuosity of vessels of the bulbar conjunctiva. Ophthalmology 115, e27–e32 (2008).

    Article  Google Scholar 

  19. Owen, C. G., Newsom, R. S. B., Rudnicka, A. R., Ellis, T. J. & Woodward, E. G. Vascular response of the bulbar conjunctiva to diabetes and elevated blood pressure. Ophthalmology 112, 1801–1808 (2005).

    Article  Google Scholar 

  20. Khan, M. A. et al. A clinical correlation of conjunctival microangiopathy with grades of retinopathy in type 2 diabetes mellitus. Armed Forces Med. J. India 73, 261–266 (2017).

    Article  CAS  Google Scholar 

  21. Sharma, R., Sati, A., Shankar, S. & Gurunadh, V. S. Use of conjunctival vessel width for assessment of severity of retinopathy in type-2 diabetes mellitus patients. Int. J. Contemp. Med. Res. 4, 1007–1010 (2017).

    Google Scholar 

  22. Chang, H.-C., Sung, C.-W. & Lin, M.-H. Serum lipids and risk of atherosclerosis in xanthelasma palpebrarum: a systematic review and meta-analysis. J. Am. Acad. Dermatol. 82, 596–605 (2020).

    Article  CAS  Google Scholar 

  23. Piyasena, M. M. P. N. et al. Systematic review on barriers and enablers for access to diabetic retinopathy screening services in different income settings. PLoS ONE 14, e0198979 (2019).

    Article  CAS  Google Scholar 

  24. Stevenson, M., Lloyd-Jones, M., Morgan, M. Y. & Wong, R. in Non-Invasive Diagnostic Assessment Tools for the Detection of Liver Fibrosis in Patients with Suspected Alcohol-Related Liver Disease: A Systematic Review and Economic Evaluation Appendix 8 (NIHR Journals Library, 2012).

  25. Bang, H. et al. Development and validation of a patient self-assessment score for diabetes risk. Ann. Intern. Med. 151, 775–783 (2009).

    Article  Google Scholar 

  26. American Diabetes Association and the Centers for Disease Control and Prevention. Prediabetes Risk Test. National Diabetes Prevention Program https://www.cdc.gov/diabetes/prevention/pdf/Prediabetes-Risk-Test-Final.pdf (2009).

  27. Zhang, X. et al. Access to health care and control of ABCs of diabetes. Diabetes Care 35, 1566–1571 (2012).

    Article  Google Scholar 

  28. Klonoff, D. C. & Schwartz, D. M. An economic analysis of interventions for diabetes. Diabetes Care 23, 390–404 (2000).

    Article  CAS  Google Scholar 

  29. Keenum, Z. et al. Patients’ adherence to recommended follow-up eye care after diabetic retinopathy screening in a publicly funded county clinic and factors associated with follow-up eye care use. JAMA Ophthalmol. 134, 1221–1228 (2016).

    Article  Google Scholar 

  30. Lu, Y. et al. Divergent perceptions of barriers to diabetic retinopathy screening among patients and care providers, Los Angeles, California, 2014-2015. Prev. Chronic Dis. 13, E140 (2016).

    Article  Google Scholar 

  31. Paz, S. H. et al. Noncompliance with vision care guidelines in Latinos with type 2 diabetes mellitus: the Los Angeles Latino Eye Study. Ophthalmology 113, 1372–1377 (2006).

    Article  Google Scholar 

  32. Legorreta, A. P., Hasan, M. M., Peters, A. L., Pelletier, K. R. & Leung, K. M. An intervention for enhancing compliance with screening recommendations for diabetic retinopathy. A bicoastal experience. Diabetes Care 20, 520–523 (1997).

    Article  CAS  Google Scholar 

  33. Nelson, R. H. Hyperlipidemia as a risk factor for cardiovascular disease. Prim. Care 40, 195–211 (2013).

    Article  Google Scholar 

  34. Karr, S. Epidemiology and management of hyperlipidemia. Am. J. Manag. Care 23, S139–S148 (2017).

    Google Scholar 

  35. Alexander, G. C. et al. Use and content of primary care office-based vs telemedicine care visits during the COVID-19 pandemic in the US. JAMA Netw. Open 3, e2021476 (2020).

    Article  Google Scholar 

  36. Jalil, M., Ferenczy, S. R. & Shields, C. L. iPhone 4s and iPhone 5s imaging of the eye. Ocul. Oncol. Pathol. 3, 49–55 (2017).

    Article  Google Scholar 

  37. Ludwig, C. A. et al. Training time and quality of smartphone-based anterior segment screening in rural India. Clin. Ophthalmol. 11, 1301–1307 (2017).

    Article  Google Scholar 

  38. Avram, R. et al. A digital biomarker of diabetes from smartphone-based vascular signals. Nat. Med. 26, 1576–1582 (2020).

    Article  CAS  Google Scholar 

  39. Santos, M. & Hofmann, R. J. Ocular manifestations of obstructive sleep apnea. J. Clin. Sleep Med. 13, 1345–1348 (2017).

    Article  Google Scholar 

  40. Cristescu Teodor, R. & Mihaltan, F. D. Eyelid laxity and sleep apnea syndrome: a review. Rom. J. Ophthalmol. 63, 2–9 (2019).

    Article  Google Scholar 

  41. Scott, I. U. & Siatkowski, M. R. Thyroid eye disease. Semin. Ophthalmol. 14, 52–61 (1999).

    Article  CAS  Google Scholar 

  42. Dutton, J. J. Anatomic considerations in thyroid eye disease. Ophthal. Plast. Reconstr. Surg. 34, S7–S12 (2018).

    Article  Google Scholar 

  43. Christoffersen, M. et al. Xanthelasmata, arcus corneae, and ischaemic vascular disease and death in general population: prospective cohort study. Br. Med. J. 343, d5497 (2011).

    Article  Google Scholar 

  44. To, W. J. et al. Real-time studies of hypertension using non-mydriatic fundus photography and computer-assisted intravital microscopy. Clin. Hemorheol. Microcirc. 53, 267–279 (2013).

    Article  CAS  Google Scholar 

  45. Mullaem, G. & Rosner, M. H. Ocular problems in the patient with end-stage renal disease. Semin. Dial. 25, 403–407 (2012).

    Article  Google Scholar 

  46. Klaassen-Broekema, N. & van Bijsterveld, O. P. Limbal and corneal calcification in patients with chronic renal failure. Br. J. Ophthalmol. 77, 569–571 (1993).

    Article  CAS  Google Scholar 

  47. Sharon, Y. & Schlesinger, N. Beyond joints: a review of ocular abnormalities in gout and hyperuricemia. Curr. Rheumatol. Rep. 18, 37 (2016).

    Article  Google Scholar 

  48. Hsiao, C.-H. et al. Association of severity of conjunctival and corneal calcification with all-cause 1-year mortality in maintenance haemodialysis patients. Nephrol. Dial. Transpl. 26, 1016–1023 (2011).

    Article  Google Scholar 

  49. Xiao, W. et al. Screening and identifying hepatobiliary diseases through deep learning using ocular images: a prospective, multicentre study. Lancet Digit. Health 3, e88–e97 (2021).

    Article  Google Scholar 

  50. Wagner, S. K. et al. Insights into systemic disease through retinal imaging-based oculomics. Transl. Vis. Sci. Technol. 9, 6 (2020).

    Article  Google Scholar 

  51. Panwar, N. et al. Fundus photography in the 21st century — a review of recent technological advances and their implications for worldwide healthcare. Telemed. J. E. Health 22, 198–208 (2016).

    Article  Google Scholar 

  52. Prokofyeva, E., Wegener, A. & Zrenner, E. Cataract prevalence and prevention in Europe: a literature review. Acta Ophthalmol. 91, 395–405 (2013).

    Article  Google Scholar 

  53. Madsen, P. H. Rubeosis of the iris and haemorrhagic glaucoma in patients with proliferative diabetic retinopathy. Br. J. Ophthalmol. 55, 368–371 (1971).

    Article  CAS  Google Scholar 

  54. Rodrigues, G. B. et al. Neovascular glaucoma: a review. Int J. Retin. Vitreous 2, 26 (2016).

    Article  Google Scholar 

  55. Zhou, Y., Zhang, Y., Shi, K. & Wang, C. Body mass index and risk of diabetic retinopathy: a meta-analysis and systematic review. Medicine 96, e6754 (2017).

    Article  Google Scholar 

  56. Babic, B., Gerke, S., Evgeniou, T. & Glenn Cohen, I. Direct-to-consumer medical machine learning and artificial intelligence applications. Nat. Mach. Intell. 3, 283–287 (2021).

    Article  Google Scholar 

  57. Cuadros, J. & Bresnick, G. EyePACS: an adaptable telemedicine system for diabetic retinopathy screening. J. Diabetes Sci. Technol. 3, 509–516 (2009).

    Article  Google Scholar 

  58. Chasan, J. E., Delaune, B., Maa, A. Y. & Lynch, M. G. Effect of a teleretinal screening program on eye care use and resources. JAMA Ophthalmol. 132, 1045–1051 (2014).

    Article  Google Scholar 

  59. Maa, A. Y. et al. Early experience with technology-based eye care services (TECS): a novel ophthalmologic telemedicine initiative. Ophthalmology 124, 539–546 (2017).

    Article  Google Scholar 

  60. Photographer Manual (EyePACS); https://www.eyepacs.org/photographer/protocol.jsp#external_photos (2008).

  61. Grading diabetic retinopathy from stereoscopic color fundus photographs–an extension of the modified Airlie House classification. ETDRS report number 10. Early Treatment Diabetic Retinopathy Study Research Group. Opthalmology 98, 786–806 (1991).

  62. EyePACS digital retinal grading protocol (EyePACS); https://www.eyepacs.org/consultant/Clinical/grading/EyePACS-DIGITAL-RETINAL-IMAGE-GRADING.pdf (2008).

  63. Bora, A. et al. Predicting risk of developing diabetic retinopathy using deep learning. Lancet Digit. Health 3, e10–e19 (2021).

    Article  Google Scholar 

  64. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J. & Wojna, Z. Rethinking the inception architecture for computer vision. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2818–2826 (IEEE, 2016).

  65. Gulshan, V. et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 316, 2402–2410 (2016).

    Article  Google Scholar 

  66. DeLong, E. R., DeLong, D. M. & Clarke-Pearson, D. L. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44, 837–845 (1988).

    Article  CAS  Google Scholar 

  67. Selvaraju, R. R. et al. Grad-CAM: visual explanations from deep networks via gradient-based localization. In 2017 IEEE International Conference on Computer Vision (ICCV) 618–626 (IEEE, 2017).

  68. Springenberg, J. T., Dosovitskiy, A., Brox, T. & Riedmiller, M. Striving for Simplicity: The All Convolutional Net. Preprint at arXiv [cs.LG] https://arxiv.org/abs/1412.6806 (2014).

  69. Sundararajan, M., Taly, A. & Yan, Q. Axiomatic attribution for deep networks. in Proceedings of the 34th International Conference on Machine Learning Vol. 70, 3319-3328 (2017).

  70. PAIR-code. https://github.com/PAIR-code/saliency PAIR-code/saliency. (2020).

Download references

Acknowledgements

This work was funded by Google LLC. We acknowledge H. Doan, Q. Duong, R. Lee and the Google Health team for software infrastructure support and data collection. We also thank T. Guo, M. McConnell, M. Howell and S. Kavusi for their feedback on the manuscript. We are grateful to the graders who labelled data for the pupil segmentation model.

Author information

Author notes

  1. These authors contributed equally: Boris Babenko, Akinori Mitani.

Authors and Affiliations

  1. Google Health, Palo Alto, CA, USA

    Boris Babenko, Akinori Mitani, Naho Kitade, Preeti Singh, Greg S. Corrado, Lily Peng, Dale R. Webster, Avinash Varadarajan, Naama Hammel & Yun Liu

  2. Artera, Mountain View, CA, USA

    Akinori Mitani

  3. Google Health via Advanced Clinical, Deerfield, IL, USA

    Ilana Traynis

  4. Department of Ophthalmology, Emory University School of Medicine, Atlanta, GA, USA

    April Y. Maa

  5. Regional Telehealth Services, Technology-based Eye Care Services (TECS) Division, Veterans Integrated Service Network (VISN) 7, Decatur, GA, USA

    April Y. Maa

  6. EyePACS Inc, Santa Cruz, CA, USA

    Jorge Cuadros

Authors
  1. Boris Babenko
    View author publications

    Search author on:PubMed Google Scholar

  2. Akinori Mitani
    View author publications

    Search author on:PubMed Google Scholar

  3. Ilana Traynis
    View author publications

    Search author on:PubMed Google Scholar

  4. Naho Kitade
    View author publications

    Search author on:PubMed Google Scholar

  5. Preeti Singh
    View author publications

    Search author on:PubMed Google Scholar

  6. April Y. Maa
    View author publications

    Search author on:PubMed Google Scholar

  7. Jorge Cuadros
    View author publications

    Search author on:PubMed Google Scholar

  8. Greg S. Corrado
    View author publications

    Search author on:PubMed Google Scholar

  9. Lily Peng
    View author publications

    Search author on:PubMed Google Scholar

  10. Dale R. Webster
    View author publications

    Search author on:PubMed Google Scholar

  11. Avinash Varadarajan
    View author publications

    Search author on:PubMed Google Scholar

  12. Naama Hammel
    View author publications

    Search author on:PubMed Google Scholar

  13. Yun Liu
    View author publications

    Search author on:PubMed Google Scholar

Contributions

B.B., A.M. and N.K. conducted the machine-learning model development, experiments and statistical analysis with input and guidance from A.V., N.H. and Y.L. B.B., A.M., I.T., N.H. and Y.L. designed the study and pre-specified the statistical analysis. I.T., N.K. and P.S. developed guidelines and managed data collection for the pupil/iris segmentation model. J.C. and A.Y.M. managed data collection and associated approvals. G.S.C., L.P. and D.R.W. obtained funding for data collection and analysis, supervised the study and provided strategic guidance. B.B., A.M., I.T., N.H. and Y.L. prepared the manuscript with input from all authors. B.B. and A.M. contributed equally, and A.V., N.H. and Y.L. contributed equally.

Corresponding authors

Correspondence to Naama Hammel or Yun Liu.

Ethics declarations

Competing interests

B.B., A.M., N.K., G.S.C., L.P., D.R.W., A.V., N.H. and Y.L. are employees of Google LLC, own Alphabet stock and are co-inventors on patents (in various stages) for machine learning using medical images. I.T. is a consultant of Google LLC. J.C. is the CEO of EyePACS Inc. A.Y.M. declares no competing interests.

Peer review

Peer review information

Nature Biomedical Engineering thanks Meindert Niemeijer and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Curves of positive predictive value (PPV) as a function of threshold for various predictions using external eye images.

a, poor sugar control (HbA1c ≥ 9%), b-c, elevated lipids (total cholesterol ≥ 240 mg dl-1 and triglycerides ≥ 200 mg dl-1), d, moderate-or-worse diabetic retinopathy (DR), e, diabetic macular edema (DME), f, vision-threatening DR (VTDR), and g, a positive control: cataract. In these plots, the x-axis indicates the percentage of patients predicted to be positive; for example 5% means the top 5% based on predicted likelihood was categorized to be “positive”, and the respective curves indicate the PPV for that threshold. The curves are truncated at the extreme end (when only 0.5% of patients are predicted positive, confidence intervals are wide) to reduce noise and improve clarity. Shaded areas indicate 95% bootstrap confidence intervals. Empty panels indicate unavailable data in validation set C and D. (*) Baseline characteristics models for validation sets A and B include self-reported age, sex, race/ethnicity and years with diabetes and were trained on the training dataset. (+) The baseline characteristics models for validation sets C and D use self-reported age and sex and were trained directly on validation sets C and D due to large differences in patient population compared to the development set. †: prespecified primary prediction tasks.

Extended Data Fig. 2 Curves of negative predictive values (NPV) as a function of threshold for various predictions using external eye images.

a, poor sugar control (HbA1c ≥ 9%), b-c, elevated lipids (total cholesterol ≥ 240 mg dl-1 and triglycerides ≥ 200 mg dl-1), d, moderate-or-worse diabetic retinopathy (DR), e, diabetic macular edema (DME), f, vision-threatening DR (VTDR), and g, a positive control: cataract. This is the NPV equivalent of Extended Data Fig. 1. The curves are truncated at the extreme end (when only 1% of patients are predicted negative, confidence intervals are wide) to reduce noise and improve clarity. Shaded areas indicate 95% bootstrap confidence intervals. Empty panels indicate unavailable data in validation set C and D. (*) Baseline characteristics models for validation sets A and B include self-reported age, sex, race/ethnicity and years with diabetes and were trained on the training dataset. (+) The baseline characteristics models for validation sets C and D use self-reported age and sex and were trained directly on validation sets C and D due to large differences in patient population compared to the development set. †: prespecified primary prediction tasks.

Extended Data Fig. 3 Receiver operating characteristic curves (ROCs) for various predictions using external eye images.

a, poor sugar control (HbA1c ≥ 9%), b-c, elevated lipids (total cholesterol ≥ 240 mg dl-1 and triglycerides ≥ 200 mg dl-1), d, moderate-or-worse diabetic retinopathy (DR), e, diabetic macular edema (DME), f, vision-threatening DR (VTDR), and g, a positive control: cataract. Sample sizes (“N” for the number of visits and “n” for the number of positive visits), area under ROC (AUCs), and the p-value for the difference are provided in Supplementary Table 1. Empty panels indicate unavailable data in validation sets C and D. (*) Baseline characteristics models for validation sets A and B include self-reported age, sex, race/ethnicity and years with diabetes and were trained on the training dataset. (+) The baseline characteristics models for validation sets C and D use self-reported age and sex and were trained directly on validation sets C and D due to large differences in patient population compared to the development set. †: prespecified primary prediction tasks.

Extended Data Fig. 4 Saliency analysis illustrating the influence of various regions of the image towards the prediction.

a-g, Figures are generated in the same manner as in Fig. 3, but with different saliency methods: Integrated Gradients on the left, and guided backpropagation on the right. h-i, Quantifying the pixel intensity in the averaged saliency heatmaps. †: prespecified primary prediction tasks.

Supplementary information

Main Supplementary Information

Supplementary note, figures and tables.

Reporting Summary

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Babenko, B., Mitani, A., Traynis, I. et al. Detection of signs of disease in external photographs of the eyes via deep learning. Nat. Biomed. Eng 6, 1370–1383 (2022). https://doi.org/10.1038/s41551-022-00867-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41551-022-00867-5

This article is cited by

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing