In the sixth article in our series, a simple eye examination shows what happens when artificial intelligence helps patients and clinicians act in time.
By: Ya Libnan – Article 6 in our series about how AI is improving our lives: Analysis
A person with diabetes can see perfectly well and still have damage developing at the back of the eye. By the time vision changes become obvious, the chance to intervene early may have narrowed. That is why regular retinal screening matters so much. Yet many patients miss the examination. A separate appointment, travel, cost, or the demands of daily life can stand between a patient and the care they need.
Artificial intelligence is helping close that gap. In some clinics, a staff member takes photographs of the retina during a regular diabetes visit. An authorized AI system analyzes the images and provides a screening result while the patient is still there. When the result suggests disease—or the images cannot be read reliably—the care team can arrange an examination by an eye specialist. The software does not treat the eye. It helps make the first step easier to complete.
A randomized study at a pediatric diabetes center affiliated with Johns Hopkins University gives us a striking example. Researchers enrolled 164 young people, ages 8 to 21, with type 1 or type 2 diabetes. Eighty-one were assigned to receive an AI eye examination at the diabetes clinic. Eighty-three received the usual referral and education about seeing an eye-care professional. Within six months, every participant in the AI group had completed an eye examination, compared with 22 percent in the referral group.
The result did not end with the camera. Twenty-five young people in the AI group had an abnormal screening result. Sixteen of them followed through with an eye-care professional. That is progress, but it also reveals the next challenge: identifying a problem is only useful when patients can obtain the care that follows.
A later Johns Hopkins analysis of adults found an association between AI screening in primary-care clinics and better access to specialist eye care for some patients. The researchers themselves cautioned that this observational finding cannot establish that AI alone caused the improvement. We should welcome the result without claiming more than the evidence shows.
For a young person with diabetes, the practical difference is easy to picture. Instead of leaving the clinic with a referral that may never become an appointment, the patient can receive the screening that day. If the result calls for specialist care, the family leaves knowing that another step is needed. AI helps the clinician start the conversation sooner; the family and medical team carry it forward.
This is a success story about access, not a claim that a computer has cured blindness. The randomized study measured completed examinations and follow-up, not how many participants ultimately avoided vision loss. Early detection and timely treatment are the path by which screening can protect sight. We need both steps, and we need to make them available to every patient who needs them.
I have lived with diabetes long enough to know how easily one complication can change a life. That is why this example matters to me. A technology that helps a child receive an eye examination during a routine visit may sound modest compared with the grand promises made about AI. To that child and family, keeping a chance to protect their vision is anything but modest.
The best future for AI is one in which it strengthens human care. A camera captures the image. AI helps assess it. A clinician explains the result. A specialist treats the patient when necessary. And a person gets a better chance to keep seeing the world.
Source notes
- Wolf and colleagues, Nature Communications (2024), ACCESS randomized trial: https://pure.johnshopkins.edu/en/publications/autonomous-artificial-intelligence-increases-screening-and-follow/
- Leong and colleagues, npj Digital Medicine (2026), exploratory adult care study: https://pmc.ncbi.nlm.nih.gov/articles/PMC13077085/
