This week's rounds
Your eye check result, before you leave.
If you have diabetes, a quick photo of the back of your eye can catch damage early. The hard part is often getting to the eye clinic afterward.
In a trial at 24 hospitals in Tanzania, 2,378 adults with diabetes had their eyes photographed. For half, AI read the photos on the spot, and a health worker gave them the result face to face. The other half got the usual path: photos read by a person later, with the result by phone a few weeks on.
Among people who needed a specialist, 35.9% reached the eye clinic within eight weeks, versus 18.2% with the usual path. That's almost twice as many, and they got there sooner: about 11 days on average versus 37, according to LSHTM.
- The win came from the quick result and the conversation together. The trial can't say how much was down to the AI alone.
- Most people who needed care still hadn't reached the clinic within eight weeks.
- One country, two regions. It measured clinic visits, not sight saved.
- Screening that refers people on, not a diagnosis.
Study: Cleland et al., "Artificial intelligence-supported diabetic retinopathy screening in Tanzania: a randomised controlled trial," The Lancet Global Health, Oct 12, 2026. Read the study · LSHTM summary
Your seizure monitor can now talk back.
A Penn team let patients text with an AI linked to their brain-wave monitor, and the seizure detector learned from their feedback in testing. An early study: 13 adults, one hospital, a research prototype.