Artificial Intelligence for DR detection in LMICs

The early diagnosis and treatment of Diabetic Retinopathy (DR) is known to preserve vision; this is done successfully in the UK through the national DR screening programme. However, there are very few DR screening programmes in Africa, and those that do exist suffer from a shortage of specialist eye care staff and high rates of loss to follow-up.

In Tanzania, we completed a clinical trial to assess whether using artificial intelligence (AI) in a real-world programme can improve DR screening outcomes for patients and help prevent people losing vision from DR.

Currently, people within the DR service in this region of Tanzania may have to wait several days or weeks for a doctor back at a hospital to analyse their retinal image and be told they have DR and require follow-up treatment. This creates time in which people are lost to services and don’t return for treatment. AI diagnosis at the point of screening can reduce this waiting time and increase the amount of people who attend care.

By automating the clinical decision making (in a validated manner), using AI also reduces the pressure on the workforce. Trained doctors at the hospital can spend less time analysing images or travelling to remote clinics, while local health workers can use the AI systems quickly and easily.

The trial included 2,378 people attending a diabetic eye screening programmes across 24 hospitals in Tanzania. Half of the participants received an AI-supported screening pathway with immediate counselling, while the other half were given the standard screening pathway.

The results showed that nearly twice as many people in the AI-enabled arm attended a follow-up appointment compared to people in the standard arm. Of those who received AI-supported screening and had a referable level of diabetic retinopathy, 35.9% attended a specialist clinic within eight weeks, compared to 18.2% of those who received the standard screening. 

Those who attended specialist care after AI screening also did so sooner, an average of 10.9 days after screening compared with 36.9 days for those receiving standard screening. The findings, published in The Lancet Global Health, suggest integrating AI into screening programmes could give healthcare workers more time to provide face-to-face support to patients, particularly in low-resource settings,for conditions where early intervention and treatment is crucial.

The difference may be partially explained by how patients received their results. In the AI-supported pathway, patients were provided with an immediate result, allowing for face-to-face counselling from healthcare workers. In the standard pathway, retinal images were assessed by a healthcare worker at a nearby hospital, and results and counselling were given weeks later over the phone.

These findings suggest integrating AI into screening programmes could help address some of these barriers by providing immediate results and creating more opportunities for patients to discuss their condition and treatment options. The findings could help to inform treatment pathways for other eye and general health conditions where an AI diagnosis and immediate counselling can improve access and efficiency.

The Tanzanian Ministry of Health, a collaborator on the study, has committed to integrating AI diagnostics into diabetic eye screening programmes in multiple regions of the country. 

Publications

Cleland C, Bascaran C, Makupa W, Shiliop B, Rwiza J, Nyale L, Macha E, Mumba F, Sandi F, Philippin H, Marques AP, Bolter L, Anderson A, Egan C, Tufail A, Keane P, Denniston A, Macleod D, Burton M. Artificial intelligence-supported diabetic retinopathy screening in Tanzania: a randomised controlled trial. The Lancet Global Health 2026. https://doi.org/10.1016/j.langlo.2026.104062

Cleland CR, Bascaran C, Makupa W, et al. Artificial intelligence-supported diabetic retinopathy screening in Tanzania: rationale and design of a randomised controlled trial. BMJ Open. 2024;14:e075055. https://dx.doi.org/10.1136/bmjopen-2023-075055

Acknowledgments

Our partners for the project are the Kilimanjaro Christian Medical Centre in Moshi and Muhimbili Hospital in Dar es Salaam. Funding is from the British Council for the Prevention of Blindness (BCPB), Christian Blind Mission (CBM) and Sir Halley Stewart Trust.