AI can aid with early detection of diabetic retinopathy
Michael Abramoff, MD, PhD, a retina specialist and founder and executive chairman of Digital Diagnostics, points out that much evidence has been collected over the past few decades showing that early detection is key.
(Image Credit: AdobeStock/vxnaghiyev)
For patients diagnosed with diabetes, early detection of diseases like diabetic retinopathy, which can rob individuals of their vision, is paramount.
According to the National Diabetes Statistics Report from the Centers for Disease Control, 38.4 million Americans have diabetes. Of those, 29.7 million people have been diagnosed, and about 8.7 million people go undiagnosed.1
Catching diabetic retinopathy early can help preserve vision, Michael Abramoff, MD, PhD, a retina specialist and founder and executive chairman of Digital Diagnostics, points out that there has been so much evidence collected over the past few decades showing that early detection is key.
“It became clear that by the time there were symptoms, like a vitreous hemorrhage or retinal detachment, the treatments available were not very effective,” he explained. “It turned out to be really important to find this disease early before there were symptoms. So that is almost decades ago.”
Abramoff pointed out that over the last few decades, ophthalmologists and optometrists as a community have worked to ensure that patients get diabetic eye exams.
“It has been about the education of patients to make sure they are taking care of their sight,” he said. “It also is about educating the physicians and nurses and endocrinologist to take care of these patients make sure that they refer these patients as part of the standard of care right from the start.”
Abramoff pointed out that recent studies from claims-based data show about 15% of individuals receive diabetic eye exams, meaning about 85% do not.2
“It hasn't really moved the needle even though we have the scientific knowledge, and from ophthalmology and optometry, we know what to do and it's still not happening,” he said.
AI is giving patients another option. The struggle had been getting the person to the eye exam, and the technology now allows the examination to go to the person.
Even with the technology in place, Abramoff noted it has been a struggle to line up approvals, reimbursement, ethical framework, and getting all stakeholders in healthcare to acknowledge the AI provided a safe, effective option for examinations.
Moreover, Abramoffnoted that another paper published in November 2023 in the
New England Journal of Medicine showed that AI diagnostics are the fastest growing AI options in terms of patient utilization.
“We now have recent randomized, controlled trials that show that not only do people get these diabetic eye exams, if you install this autonomous AI, it makes health disparities almost disappear,” he said.
The autonomous AI examination option increased compliance among minority and rural populations. With early detection important to saving vision in many eye diseases, Abramoff noted the AI diagnostics can help prevent blindness in patients.
How it works
Abramoff noted that the AI makes it diagnosis by comparing the patient’s images to a repository of high-quality images. The AI exam can be done in the office of the endocrinologist, and a referral can be made to an ophthalmologist or retina specialist if necessary.
Patients with diabetes visit their endocrinologist and receive an AI vision exam. It could detect the early signs of diabetic retinopathy, and they are then referred to an ophthalmologist or retina specialist.
The accuracy of the AI diagnosis is high, according to Abramoff, himself a retina specialist.
“There are times when I disagree with myself,” he pointed out. “I will see a patient 2 weeks later. Nothing has changed but I may make a change based on what I have seen in other patients.”
Abramoff also pointed out that autonomous AI is reaching populations of patients that are varied from those that clinicians usually see.
“I think more evidence will be coming out, and that will be a big, big thing for ophthalmology,” he said.
Another key is reimbursement, and integrating AI technologies in healthcare settings that have national reimbursement to support the adoption of such technology.
“Without sustainable reimbursement wide-spread adoption of healthcare AI wouldn’t be possible,” Abramoff said. “There is a ton of noise out there about AI, and we need to stop talking about it in terms of potential, as the scientific evidence is now in and actionable; we need to focus on adoption at scale because people’s vision and lives are at stake.”
The future of autonomous AI
Going forward, Abramoff said he anticipates a rapid expansion of the technology with even more integration into the workflow in underserved communities.
Once the patients are diagnosed, AI can also serve as a tool to manage their care and provide vision-saving care from ophthalmologists and retinal specialists.
“We're getting all these extra patients now that had no access to care,” Abramoff said. “Now that they have the disease diagnosed, we need to be able to treat them and see them.”
For many patients, the AI diagnosis is smooth and transparent and it will help those who are diagnosed with diabetic retinopathy receive sight-saving care.
“As ophthalmologists, we will see more interesting patients and provide the care they need,” Abramoff said. “We can see more patients than would be possible without AI.”
References:
Centers for Disease Control and Prevention. National Diabetes Statistics Report. Diabetes. Published May 13, 2024. https://www.cdc.gov/diabetes/php/data-research/index.html
Channa R, Wolf R, Abramoff MD. Autonomous Artificial Intelligence in Diabetic Retinopathy: From Algorithm to Clinical Application. J Diabetes Sci Technol. 2021 May;15(3):695-698. doi: 10.1177/1932296820909900. Epub 2020 Mar 4. PMID: 32126819; PMCID: PMC8120059.
