icon-arrow-back-perspectiveicon-arrow-forward-perspectiveicon-arrow-nexticon-arrow-next-nofillicon-arrow-previousicon-arrow-previousicon-closegoogle-plusinstagramlinkedinnavigate-nextnavigate-previousicon-social-closeicon-social-expandtwitter Skip to main content
    A human eye and imagery that represents artificial intelligence and a retinal scan

    Hans Kaspersetz

    President

    April 25, 2018

    April Showers Bring May Flowers and an AI Medical Device FDA Approval

    Healthcare systems around the world are struggling with the economic and social impact of diabetes. In 2014, there were 422 million people being treated for diabetes. Between 1980 and 2014, the age-adjusted global prevalence of diabetes increased from 4.7% to 8.5% — an 81% increase.1 

    Diabetes causes multiple long-term complications, including vision loss. Diabetic retinopathy (DR) is the leading cause of vision loss in adults aged 20 to 74 years.2 In 2012, approximately 93 million people were living with DR.3 The elevated levels of glucose caused by diabetes damage blood vessels, causing blood and other fluids to leak. When the leaked fluid accumulates in the retina, vision is blurred. Poor circulation in the retinal blood vessels also may produce fragile new blood vessels.4 Fragile blood vessels may grow over the retina and rupture, leaking blood that blurs vision. Fragile vasculature may also rob vision by producing scar tissue that causes the retina to detach from the back of the eye. 

    The good news is that vision loss and blindness caused by DR is preventable with early intervention. The bad news is that only about half of the at-risk population pursues the annual examination from a specialist (ophthalmologist or optometrist) who would detect DR.5 

    “Early detection of retinopathy is an important part of managing care for the millions of people with diabetes, yet many patients with diabetes are not adequately screened for diabetic retinopathy.”5

    Malvina Eydelman, M.D., 
    Director of the Division of Ophthalmic, and Ear, Nose and Throat Devices 
    FDA Center for Devices and Radiological Health

    Using AI to Increase Access to Testing for Diabetic Retinopathy

    On April 11, 2018, the FDA approved an artificial intelligence algorithm that enables general practitioners (GPs) to perform examinations for diabetic retinopathy and deliver an accurate diagnosis.5 The expectation is that more people with diabetes will be tested for DR if they can be examined during visits with their general practitioners.

    The AI system has 2 components:

    • A retinal camera to capture images of the patient’s retina
    • IDx-DR software: an artificial intelligence algorithm that analyzes the retinal images

    The GP uses the camera to capture images of the patient’s retinas and uploads the image files to a cloud server running the IDx-DR software, which delivers one of two diagnoses:

    1. “More than mild diabetic retinopathy detected: refer to an eye care professional.” 
    2. “Negative for more than mild diabetic retinopathy; rescreen in 12 months.” 

    Patients who receive a positive diagnosis are referred to an eye specialist for immediate follow-up.

    Image of an eye and copy stating that diabetic retinopathy is the leading cause of vision loss in adults age 20 to age 74

    Ramifications Across Healthcare 

    By providing GPs with augmented diagnostic capabilities normally reserved for an ophthalmologist, IDx-DR moves specialized care closer to the people who need it. 

    Shifting some forms of specialized care away from hospitals and specialists and into the community is frequently proposed as a strategy to improve the effectiveness and reduce the cost of healthcare.6,7 Although data suggest that moving care into the community improves patient satisfaction, evidence is not conclusive that this approach is cost effective.8 

    One study in the UK evaluated 27 initiatives by the National Health Service (NHS) to move healthcare into the community.8 In the majority of initiatives, moving care closer to the patient improved patient care, but failed to save money for the NHS. However, 7 initiatives also reduced cost, including one study that gave GPs better access to specialist expertise — exactly what the IDx-DR system accomplishes.8 

    More Eyes on More Prizes

    General practitioners with diabetic patients might be the first group to benefit from AI technology, but they won’t be the last. Researchers at Stanford University School of Medicine have teamed with Google to develop AI technology — also based on retinal scans — that identifies cardiovascular risk factors AND predicts cardiovascular events.9,10 The scans distinguished smokers from nonsmokers and predicted systolic blood pressure within 11 mm Hg, on average, for patients overall, whether or not they suffered from high blood pressure. When the AI solution evaluated retrospective data, it predicted 70% of the patients who eventually had a cardiovascular event, including one event that occurred 5 years after the retinal scan.9 

    AI is being leveraged in many ways to enhance healthcare delivery.11 The FDA, physicians, healthcare economists, and the insurance industry will be monitoring the impact of IDx-DR and similar AI-augmented devices on the quality and cost of healthcare delivery. Imaging and diagnostics are likely to be among the first medical specialties disrupted by AI because of the algorithm’s capability to identify weak but important signals and tirelessly process large volumes of data. Furthermore, there are significant sets of annotated scans and images that can be used to train the machine learning algorithms to detect pathologies.

    As technologists who have leveraged machine learning and AI to uncover customer behavior and enhance SEO and content strategy and as people whose lives have been touched by diabetes, we’ll be watching, too. We’ll keep you informed.

    References

    1. World Health Organization. Global Report on Diabetes. Geneva, Switzerland: World Health Organization, 2016. http://apps.who.int/iris/bitstream/handle/10665/204871/9789241565257_eng.pdf;jsessionid=8E1FA5218330AA20AAAD26065B07BE31?sequence=1. Accessed April 25, 2018.
    2. Lee R, Wong TY, Sabanayagam C. Epidemiology of diabetic retinopathy, diabetic macular edema and related vision loss. Eye Vis (Lond). 2015:2:17. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4657234/. Accessed April 25, 2018.
    3. Yau JWY, Rogers SL, Kawasaki R, et al. Global prevalence and major risk factors of diabetic retinopathy. Diabetes Care. 2012;25(3):556-564. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3322721/. Accessed April 25, 2018.
    4. Diabetic retinopathy: your questions answered. Joslin Diabetes Center website. https://www.joslin.org/info/Diabetic_Retinopathy_What_You_Need_to_Know.html. Accessed April 24, 2018.
    5. Food and Drug Administration. FDA permits marketing of artificial intelligence-based device to detect certain diabetes-related eye problems [press release]. FDA website. April 11, 2018. https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm604357.htm.
    6. Miani C, Winpenny E. Moving outpatient care into the community. The Rand Blog. December 8, 2016. https://www.rand.org/blog/2016/12/moving-outpatient-care-into-the-community.html. Accessed April 25, 2018.
    7. Winpenny E, Miani C, Pitchforth E, et al. Outpatient services and primary care: scoping review, substudies and international comparisons. Health Serv Deliv Res. 2016;4(15):1-322. https://www.rand.org/pubs/external_publications/EP66473.html. Accessed April 25, 2018.
    8. Moving care out of hospital won't save money, experts warn. Nuffield Trust website. January 3, 2017. https://www.nuffieldtrust.org.uk/news-item/moving-care-out-of-hospital-won-t-save-money-experts-warn. Accessed April 25, 2018.
    9. Poplin R, Varadarajan AV, Blumer K, et al. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nature Biomed Eng. 2018;2:158-164. https://www.nature.com/articles/s41551-018-0195-0. Accessed April 25, 2018.
    10. Peng L. Assessing cardiovascular risk factors with computer vision. Google Research Blog. https://research.googleblog.com/2018/02/assessing-cardiovascular-risk-factors.html. February 19, 2018. Accessed April 24, 2018.
    11. Faggella D. Machine learning healthcare applications—2018 and beyond. Techemergence website. https://www.techemergence.com/machine-learning-healthcare-applications. Updated March 22, 2018. Accessed April 24, 2018.