LumineticsCore® is highlighted in a February 1, 2026 American Academy of Ophthalmology (AAO) article about AI-powered solutions expanding access to diabetic retinopathy exams by bringing high-quality diagnostics into primary care. In the article, Digital Diagnostics’ Founder and Executive Chairman Michael D. Abramoff, MD, PhD, notes: “The point is, these autonomous AI systems are always meant to be where the eye specialist cannot be, but where the patient is.” The AAO article is provided via link on the Digital Diagnostics web site and reproduced below.
Artificial intelligence (AI) has rapidly evolved from a promising idea to a practical clinical tool in medicine. The field of ophthalmology has made significant steps forward in the adoption of AI technology, especially for diabetic retinopathy (DR) screening.1
Globally, 103.12 million adults were estimated to be living with DR in 2020.2 In the United States, about 9.6 million people have the condition.3 But many people remain unscreened each year due to barriers to access, financial challenges, and limited health care infrastructures where they live.
AI-driven screening solutions are emerging as a way to expand access to DR screening, especially in primary care settings and in under-resourced areas.4,5
FDA-Approved AI Systems for DR Screening
Currently, three autonomous AI systems have been approved by the U.S. FDA for the detection of DR. Autonomous systems do not require human oversight to make a clinical decision and give diagnostic results. (AI-assisted or AI-augmented systems, by comparison, are limited to analyzing data and providing suggestions to a human clinician who makes diagnostic decisions.)
The three systems include (in order of FDA-approval):
LumineticsCore (Digital Diagnostics). In 2018, this was the first AI system de novo cleared by the FDA for DR screening. It has also received the first clearance for an autonomous AI algorithm.
EyeArt System (EyeNuk). EyeArt was approved in 2020. In addition to autonomous AI DR detection, the system offers an AI-assisted mode, which can support ophthalmologists who review images for glaucoma and age-related macular degeneration.
AEYE-DS (AEYE Health). The initial version of AEYE-DS, which used a tabletop camera, was approved in 2022.
These systems capture digital fundus photographs (typically nonmydriatic unless dilation is needed for image quality), upload them to cloud-based software, and generate an automated assessment indicating whether more-than-mild DR or vision-threatening DR is present, said Michael D. Abramoff, MD, PhD, Professor of Ophthalmology at the University of Iowa, Carver College of Medicine, in Iowa City, and founder of Digital Diagnostics.
Judy E. Kim, MD, the Jean and Tom Walter Distinguished Chair in Ophthalmology in Honor of James P. McCulley MD at UT Southwestern Medical Center, in Dallas, noted that the AEYE-DS system also received FDA clearance for use with the Optomed Aurora portable handheld camera.
Kim said, “Multiple other research and commercial algorithms are in development globally and more predicate 510(k) filings have appeared in 2023-2025.”
The AI tools can be utilized in primary care settings during routine checkups for patients with diabetes. Results are provided within a minute or so, indicating whether the patient should be further evaluated by an ophthalmologist or wait to be screened again in a year, Abramoff said.
Benefits of AI Systems for DR Detection
There are several benefits of autonomous AI tools for DR screening, for both ophthalmologists and patients, experts say, including:
Accuracy. In clinical trials, these systems have shown high rates of sensitivity and specificity. For example, a study of 900 fully screened patients, which led to FDA clearance for LumineticsCore (formerly called IDx-DR), showed 87% sensitivity, with 90% specificity for more-than-mild DR.6
In a study of 942 individuals with diabetes, EyeArt’s autonomous AI mode showed 96% sensitivity and 88% specificity for detecting eyes with more-than-mild DR and 97% sensitivity and 90% specificity for vision-threatening DR. The system accurately graded more than 97% of manually scored eyes, with dilation needed only in a minority of cases.7
In clinical trials, AEYE Health reported a sensitivity of 93% and a specificity of 91.4% on desktop cameras, and 92% sensitivity and 93.6% specificity on handheld cameras. The system uses a single image per eye to generate a result, and less than 1% of patients require dilation.4,8
“The reality is, with how accurate these systems are now, they’re only going to get better and better in the future,” said Paula Ko, MD, a retina specialist with Eye Physicians and Surgeons, in Wilmington, Delaware, and Associate Clinical Professor of Ophthalmology at Temple University, in Philadelphia. “That’s how AI works. The more data you have, the more accurate it is.”
Convenience. “The problem we’re trying to solve is that way too many people are going blind from diabetes,” said Abramoff. “We know exactly how to prevent it, but we need to catch it early before patients start showing symptoms. But too many people aren’t making that separate eye appointment for an annual DR screening.”
Six of 10 people with diabetes skip their annual eye exam, he said. Abramoff noted that barriers can include transportation, scheduling around work or other commitments, childcare, costs and insurance, and referral complexity. People with diabetes may also not understand the risks of DR, especially if symptoms aren’t present yet, he said.
If a primary care physician or endocrinologist is seeing the patient for their regular diabetes appointment, an AI system allows them to do the DR screening at the same time. “The point is, these autonomous AI systems are always meant to be where the eye specialist cannot be, but where the patient is,” Abramoff said.
User-Friendliness. “In my estimation, it probably takes no more than 30 minutes to train someone to operate a camera and take the photos accordingly,” said Tin Yan Alvin Liu, MD, Endowed Professor in Ophthalmology and AI, and Founding Director, Gills AI Innovation Center at Johns Hopkins University, in Baltimore. “Really, any nonmedical personnel could be trained to use one.”
Speed. An AI system’s ability to quickly generate a diagnosis has a positive impact on patients, said Sally L. Baxter, MD, MSc, Associate Professor and Division Chief for Ophthalmology Informatics and Data Science at the Shiley Eye Institute at the University of California, in San Diego.
“For patients, having an immediate imaging interpretation can give them critical information at the point of care, enhancing their likelihood to follow up with an in-person eye exam and achieve earlier diagnosis and treatment for this vision-threatening condition,” said Baxter.
Ko works with an Academy-funded pilot program for DR screening at Henrietta Johnson Medical Center, a federally qualified health center (FQHC), in Wilmington, Delaware. The program works in conjunction with EyeCare America using EyeNuk’s EyeArt AI system. “Within seconds, patients have a diagnosis if they have retinopathy or they learn that they don’t have it. That is very meaningful to them,” she said, and the instant information helps motivate them to follow up with an ophthalmologist. (The Academy is looking for EyeCare America volunteers, especially retina specialists. To learn more, visit aao.org/eyecare-america/volunteer-ophthalmologists.)
The immediacy of the diagnosis also helps physicians who are doing the AI screenings know which cases are the most serious and require prioritization, Ko said.
Productivity. AI screening tools also promote productivity among physicians and within the health care system, Abramoff said. He and colleagues conducted a randomized controlled trial where they observed 105 clinic days, 51 of which used AI diagnosis in 494 patients, and 54 of which did not use AI tools in 499 patients. They reported that AI led to 40% higher productivity, as measured by the number of completed patient encounters per hour per specialist.9
“The demand for high-quality health care is infinite, but there’s only a limited amount of humans who can provide health care,” said Abramoff. “So, I think AI is the only solution.”
Access. Places with limited access to ophthalmologists will benefit from AI systems for DR screening. “This will be especially good for rural regions,” Ko said.
“If you live two hours away from an eye doctor or someone who can do diabetic retinopathy eye screening, you’re not going to drive two hours if you think your vision is fine,” she said. “But if you get tested at your primary care doctor’s office and you know in seconds that you have retinopathy, you’ll be much more likely to get to that ophthalmologist.”
Ko said FQHCs are ideal locations for autonomous AI testing for DR because patients who get medical care at FQHCs are also typically less likely to get screened at an eye doctor’s office. They tend to be in a lower socioeconomic bracket and have poor access to health care. They also often do not have insurance or they’re underinsured. “They are people in the care gap,” she said.
Reimbursement. Autonomous AI eye examinations may be reimbursable under CPT code 92229, which was added in 2020 by the American Medical Association. The national average Medicare payment amount was $47.06 in 2022, $45.75 in 2023, and $40.94 in 2024. In 2025, it was $43.67. Many commercial payers have also followed suit, offering reimbursement for autonomous AI eye tests.4,10,11
“Broadly speaking, we do have good buy-in from the payers, with CMS and commercial payers, with national coverage,” Liu said.
Barriers to Implementing AI Screening for DR
“Adoption has been incremental but real,” said Kim, with several large health systems and primary-care networks—academic and community—having implemented autonomous DR screening pilots or programs. But she said it is not yet universal. “Many practices are still evaluating integration, return on investment, and EHR interoperability,” said Kim.
The experts described some of the challenges and pitfalls to implementing these systems in both large institutions and smaller practices.
Stakeholder and Physician Acceptance. To successfully implement AI into a DR screening program, a number of stakeholders are required, including primary care physicians, medical assistants, administrators, ophthalmologists, and patients, Liu said.
“You need to make sure that all the stakeholders align and are incentivized to make it work for an extended period of time. Just like most things in life, aligning stakeholder expectations and having their sustainable buy-in is not easy, especially with a new technology like AI that requires a somewhat different workflow,” Liu added.
It can be particularly difficult in a large health care system or academic medical center, Ko noted.
Ko said it is essential that leadership gets primary care physicians on board in bigger health systems because they’re the ones, in many cases, who will be instituting the use of new AI systems, training their staff to use the tools, and working with patients and ophthalmologists.
Smaller medical practices, where communication is often more streamlined, might have an advantage over larger institutions, Ko said. For example, she noted that for the pilot program at Henrietta Johnson, the team got the AI screening program up and running quickly. “They were small and nimble enough to make the workflow changes that were needed,” she said. “It can be a lot harder to implement change at a huge health care institution.”
Tech Integration. Information technology integration can be another hill to climb, particularly for larger institutions, Baxter said.
“We want to make sure that the information and output from the AI tool is seamlessly connected to the other routine operational aspects of our health IT systems,” she said. “This is easier said than done, in many cases. It doesn’t happen automatically and often requires substantial investment in enterprise-level IT support and infrastructure building.”
Kim said, “Practical barriers include hardware compatibility—only specific camera models are FDA-cleared for each algorithm—and integrating results into EHRs and referral workflows, staff training, image quality and ungradable rates, and upfront costs or subscription pricing.”
High Turnover in Health Care Offices. While the AI platforms to detect DR are relatively simple to use, high turnover in the health care field can make it a challenge to keep AI-trained technicians on staff.
“With the current labor market, and certainly in general, there’s relatively high turnover among some medical support staff roles,” Liu said.
Ko has seen this play out, as well. “You can train three or four people, and half of them are gone six months later,” she said. “I think best practice is that you need at least one nondoctor who really knows how to use the system, and if she’s not there, then there should be people who can fill in, but you really need that one person.”
Limitations of the Tools and Processes Around Their Use. Some of the AI products approved by the FDA to screen for DR only use desktop-based cameras. “These cameras take good photos, but the problem is that they’re still relatively bulky and fairly immobile,” Liu said. “If we were to scale this kind of technology in rural areas, we would really need more portable cameras.”
But portable cameras come with challenges, too. “Because of physics, smaller cameras typically take worse quality images, which affects AI performance,” Liu said. “So, there is some physical limit in how scalable it is when it comes to pairing AI with portable cameras. That’s the next frontier.”
Another issue for more remote areas, Baxter noted, is that even if a medical practice can successfully screen for DR using AI, it may not necessarily have the ability to take on the patient who screens positive, or that patient may not be able to access an ophthalmologist in their region. “So, an increase in screening rates may not necessarily correspond to an increase in treatment rates,” she said.
Also, the AI systems are limited in what they can do. Those that are currently approved by the FDA to screen for DR can only detect the one condition autonomously.
“If you’re using an AI model that has a binary classification like DR versus no DR, the model isn’t looking for other eye pathology and those diagnoses may be missed,” Baxter said.
Kim said liability and medicolegal questions (such as who is responsible for missed disease other than DR), patient consent and communication, and how results are documented are still evolving. “Robust referral pathways for positive screens are essential; otherwise, screening without reliable follow-up delivers limited benefit,” she said.
Fear of Change. In addition to logistical and technical barriers, there is the very real human emotion of fear. Whether it’s mistrust of new technology or worries about job loss, hesitancy around change is common when a paradigm shift is introduced, said Abramoff.
In 2010, a colleague dubbed Abramoff the “Retinator” (a play on the science fiction film The Terminator) and questioned whether the day might soon come when robots would replace ophthalmologists. “The thinking was that this will result in job loss. It will lower the quality of care, and should we really want this?” Abramoff said.
But rather than taking this as a setback, Abramoff saw it as an opportunity to more clearly shape the messaging around AI’s role in ophthalmology. “I realized that the criticisms had nothing to do with the science but everything to do with fear,” he said. “What AI allows us to do is serve every patient and then with our high expertise and high skill level, take care of those patients who need those skills.”
The Future
Ko envisions a future where autonomous AI screening is widely available, not only at primary care offices but also at health clinics and pharmacies. She pointed out that this type of access to early screening could save money and eyesight significantly on the back end. “We have amazing treatments for severe proliferative diabetic retinopathy, but 99% of that is preventable with early screening,” she said.
The tools could also be adopted to help diagnose other eye diseases like glaucoma, macular degeneration, and diabetic macular edema, she said.
As AI is used with greater frequency, Baxter believes the next natural step will be for its increasing use in diagnosis and management in ophthalmology practices, as well. “I think we will see AI assistance with recommending intravitreal injection protocols, automated interpretations of images, and generating care plans for individual patients,” she said. “Another potential use of AI will be for population health management, to monitor who is due for a follow-up or injection and who is overdue. And AI could automate outreach and phone calls to bring those patients back in.”
___________________________
1 Wolf RM et al. Nat Commun. 2024;15(1):421.
2 Teo ZL et al. Ophthalmology. 2021;128(11):1580-1591.
3 Lundeen EA et al. JAMA Ophthalmol. 2023;141(8):747-754.
4 Teng CW et al. Ophthalmol Sci. 2026;6(1);100935.
5 Yim D et al. Indian J Ophthalmol. 2021;69(11):2987-2992.
6 Abramoff MD et al. NPJ Digit Med. 2018;1(39).
7 Ipp E et al. JAMA Netw Open. 2021;4(11):e2134254.
8 Clinicaltrials.gov. https://clinicaltrials.gov/study/NCT04612868. Accessed Dec. 1, 2025.
9 Abramoff MD et al. NPJ Digit Med. 2023;6(1):184.
10 AMA. www.ama-assn.org/press-center/ama-press-releases/ama-releases-2021-cpt-code-set. Accessed Dec. 3, 2025.
11 CMS. www.cms.gov/medicare/physician-fee-schedule/search. Accessed Dec. 3, 2025.
Meet the Experts
Michael D. Abramoff, MD, PhD Professor of Ophthalmology at the University of Iowa, Carver College of Medicine, in Iowa City, and Founder of Digital Diagnostics. Relevant financial disclosures: Collaborative Community on Ophthalmic Imaging: EE; Digital Diagnostics: P,C,EE,PS; Healthcare AI Coalition: EE.
Sally L. Baxter, MD, MSc Associate Professor and Division Chief for Ophthalmology Informatics and Data Science at the Shiley Eye Institute at the University of California San Diego. Relevant financial disclosures: None.
Judy E. Kim, MD Professor of Ophthalmology, Vice-Chair of Education, Medical Director of Clinical Research, Jean and Tom Walter Distinguished Chair in Ophthalmology in Honor of James P. McCulley MD at UT Southwestern Medical Center, in Dallas. Relevant financial disclosures: None.
Paula Ko, MD Retina specialist at Eye Physicians and Surgeons, in Wilmington, Delaware, and Associate Clinical Professor of Ophthalmology at Temple University, in Philadelphia. Relevant financial disclosures: None.
Tin Yan Alvin Liu, MD Endowed Professor in Ophthalmology and AI, and Founding Director, Gills AI Innovation Center at Johns Hopkins Medicine, in Baltimore. Relevant financial disclosures: None.
Full Financial Disclosures
Abramoff Collaborative Community on Ophthalmic Imaging: EE; Digital Diagnostics: P,C,EE,PS; Healthcare AI Coalition: EE.
Baxter NIH: S; Research to Prevent Blindness: S; The Glaucoma Foundation: S; Topcon Medical Systems: C.
Kim Adverum: C; Alcon: C; Apellis: C; Astellas: C; Bausch + Lomb: C; Coherus: C; DORC: C; EyePoint: C; Genentech: C; Hoffmann-La Roche: C; Neurotech: C; Regeneron: C.
Ko None.
Liu None.
| Disclosure Category | Code | Description |
| Consultant/Advisor | C | Consultant fee, paid advisory boards, or fees for attending a meeting. |
| Employee | E | Hired to work for compensation or received a W2 from a company. |
| Employee, executive role | EE | Hired to work in an executive role for compensation or received a W2 from a company. |
| Owner of company | EO | Ownership or controlling interest in a company, other than stock. |
| Independent contractor | I | Contracted work, including contracted research. |
| Lecture fees/Speakers bureau | L | Lecture fees or honoraria, travel fees or reimbursements when speaking at the invitation of a commercial company. |
| Patents/Royalty | P | Beneficiary of patents and/or royalties for intellectual property. |
| Equity/Stock/Stock options holder, private corporation | PS | Equity ownership, stock and/or stock options in privately owned firms, excluding mutual funds. |
| Grant support | S | Grant support or other financial support from all sources, including research support from government agencies (e.g., NIH), foundations, device manufacturers, and\or pharmaceutical companies. Research funding should be disclosed by the principal or named investigator even if your institution receives the grant and manages the funds. |
| Stock options, public or private corporation | SO | Stock options in a public or private company. |
| Equity/Stock holder, public corporation | US | Equity ownership or stock in publicly traded firms, excluding mutual funds (listed on the stock exchange) |
This posting is based on American Academy of Ophthalmology (AAO) article authored by Ashley Welch.