Back to Bigger Picture

BIGGER PICTURE Earlier diagnosis

The eye check that patients no longer had to book separately

Sometimes the breakthrough is not a more exotic test. It is completing a necessary examination before a family leaves the clinic. ACCESS tested that idea in young people with diabetes.

AI-generated illustration: of a retinal scan connected to a same-visit eye assessment.
Retinal imaging can bring an eye assessment into an existing diabetes appointment.AI-generated illustration of the idea, not a photograph of the actual people, equipment or event.
AI group screened
81 of 81
Standard referral completed
22% within six months
Examination time
5–10 minutes

How the idea works

  1. Photograph the retina

    Acquire four non-dilated eye images.

  2. Return a same-visit result

    AI checks for diabetic retinal disease.

  3. Complete or refer

    Finish screening and arrange specialist review when indicated.

A referral still asks a family to do more

Diabetic eye screening often means arranging another visit with an eye specialist. Transport, time away from school or work and uncertainty about its importance can prevent completion. Researchers at Johns Hopkins therefore studied a service problem: could screening happen during a diabetes appointment the patient was already attending?

Sources: [1]

Four photographs, a result during the visit

The system photographed the backs of the eyes without dilating drops. An algorithm analysed the images for diabetic retinal disease and returned a result during the visit. A positive finding triggered referral for further eye care. The camera therefore replaced an initial screening journey, not the specialist treatment that an abnormal finding might require.

Sources: [1][2]

Completion is the outcome

The randomised trial enrolled 164 young people aged eight to 21. All 81 offered the on-site AI examination completed it that day. In the standard-referral group, 22% obtained an eye examination within six months. This measures whether care actually happened, rather than merely whether a classifier could recognise an image.

Sources: [2]

Screening opens a door; follow-up still matters

Twenty-five AI examinations produced a positive result requiring follow-up. Such results are not equivalent to 25 confirmed cases of sight-threatening disease. The study also did not measure prevented blindness. At the time, this pediatric use was investigational. Its convincing lesson concerns access: a completed first step is more useful than an unfulfilled instruction.

Sources: [1][2]

The original sources

  1. Johns Hopkins MedicineStudy Finds AI–Driven Eye Exams Increase Screening Rates for Youth with Diabetes
  2. Nature CommunicationsAutonomous artificial intelligence increases screening and follow-up for diabetic retinopathy in youth

What to explore next

A routine ECG offered a clue to heart weakness nobody had diagnosedWhy imagining a complicated letter beat pointing at a key
Explore all 99 applications