Skip to content

Diabetic eye disease, is an important indicator of cardiovascular disease complications of diabetes, and a major cause of blindness in the working-age population.  Early detection through screening programmes can avoid further life-threatening complications and sight loss.1

The National Health Service (NHS) Diabetic Eye Screening Programme (DESP) launched in 2003, involves yearly screening using digital photographs of the retina, which are assessed by human graders for the presence of diabetic eye disease, but with 2.2 million eye screening appointments each year in England alone, generating approximately 10 million retinal images, there is an urgent need to seek computerised solutions.  

Artificial Intelligence (AI) systems designed to detect diabetic eye disease have been developed over the last 15 years and we have previously shown that they can meet pre-defined safety standards of performance and could be used to partially replace human grading.2-4  With the acceleration of AI implementation into healthcare settings, there is a need to ensure that AI-systems designed to detect presence of diabetic eye disease perform equally well for all people with diabetes.  

We will evaluate the accuracy of AI-systems to successfully triage patients into low- and medium-/high-risk diabetic eye disease cases across subgroups of the population.  We will also evaluate the perceptions, acceptability and expectations of health care professionals and people living with diabetes in relation to AI technology implementation into the eye screening programme, in particular to understand the potential impact of any ethnic differences that might lead to inequalities.

We are working with NHS AI Lab to develop a pipeline for long-term, ongoing, prospective evaluations of the best commercially available systems leading to implementation within the English NHS DESP. Our multidisciplinary teams have expertise in this area, are independent of any commercial partners and have no conflicts of interest to declare in this regard. Crucially, the project will develop safeguarding systems to ensure it works for all and that AI performance does not vary across population sub-groups, such as ethnicity or gender.

The project will also provide evidence to support the commissioning and deployment of the first potential widespread use of AI within the NHS.

1. Mohamed Q, Gillies MC, Wong TY. Management of diabetic retinopathy: a systematic review. JAMA 2007;298(8):902-16. doi: 10.1001/jama.298.8.902

2. Tufail A, Kapetanakis VV, Salas-Vega S, et al. An observational study to assess if automated diabetic retinopathy image assessment software can replace one or more steps of manual imaging grading and to determine their cost-effectiveness. Health Technol Assess 2016;20(92):1-72. doi: 10.3310/hta20920 [doi]

3. Tufail A, Rudisill C, Egan C, et al. Automated Diabetic Retinopathy Image Assessment Software: Diagnostic Accuracy and Cost-Effectiveness Compared with Human Graders. Ophthalmology 2017;124(3):343-51. doi: 10.1016/j.ophtha.2016.11.014

4. Heydon P, Egan C, Bolter L, et al. Prospective evaluation of an artificial intelligence-enabled algorithm for automated diabetic retinopathy screening of 30 000 patients. Br J Ophthalmol 2020 doi: 10.1136/bjophthalmol-2020-316594

Find a profileSearch by A-Z