1) If you had a crystal ball, how do you think using artificial intelligence (AI) technologies will change the way we prevent, detect and treat diabetic retinopathy by 2030?
AI technologies will significantly change concepts and traditional ways of management of diabetic retinopathy in the following areas:
First, AI technologies play a significant role in the screening and detecting of diabetic retinopathy based on already very mature AI algorithms, coupled with teleophthalmology platforms and mobile fundus cameras. This will be used in primary care/community settings and will increase accessibility and lower the cost of screening.
Second, AI technologies will be used to predict the onset of diabetic retinopathy. It is known that only 20-30% of patients who develop diabetic retinopathy, allowing primary care physicians to target this group for more intensive systematic therapy.
Third, AI technologies will be used to predict diabetic retinopathy progression among patients who will more likely progress from no or mild diabetic retinopathy to severe vision-threatening stages of diabetic retinopathy, including diabetic macular edema (DME). This will be used as the foundation for a new classification of diabetic retinopathy (based on AI and not on human assessment). This new classification will have prognostic value and can be used by optometrists or ophthalmologists to monitor patients more closely, and perhaps refer them to retinal specialists.
Fourth, AI technologies will be used to assist ophthalmologists in the treatment of diabetic retinopathyand DME, whether it is anti-VEGF therapy or other novel therapies. AI technologies will be applied to retinal images (fundus photos, OCT, OCT-angiography) to provide information and estimates of the indications for treatment, duration of treatment, amount of treatment needed, and longer-term outcomes.
Finally, AI technologies will be used to predict systemic complications of diabetes from eye images. Thus, a patient with diabetes could have an eye scan (fundus photo) and have information on not only the presence and risk of diabetic retinopathy but also estimates of their risk of diabetic kidney disease (DKD) and cardiovascular disease (CVD).
Reference: Sabanayagam C, et al. A deep learning algorithm to detect chronic kidney disease from retinal photographs in community-based populations. Lancet Digit Health. 2020 Jun;2(6):e295-e302.
2) Will AI technologies in diabetic retinopathy increase the access and affordability to appropriate services to underserved populations? If yes, how will it help increase access?
Yes, of course. AI technologies will revolutionise the screening of diabetic retinopathy based on mature AI algorithms coupled with teleophthalmology platforms and mobile fundus cameras, which will be used in underserved populations. Currently, such populations have no diabetic retinopathy screening programs, despite its clear value. We have demonstrated this in Zambia.
Reference: Bellemo V, et al. Artificial intelligence using deep learning to screen for referable and vision-threatening diabetic retinopathy in Africa: a clinical validation study. Lancet Digit Health. 2019 May;1(1):e35-e44
Image: Consulta oftalmologica para niños/ Javier Santayana
For Operation Eyesight, starting conversations with the communities where we work has been a key step in improving access to eye health services for women and girls. It has also helped us build gender equality when it comes to accessing local health services more generally.