Background Radical measures are required to identify and reduce blindness due to diabetes to achieve the Sustainable Development Goals by 2030. Therefore, we evaluated the accuracy of an artificial intelligence (AI) model using deep learning in a population-based diabetic retinopathy screening programme in Zambia, a lower-middle-income country.Methods We adopted an ensemble AI model consisting of a combination of two convolutional neural networks (an adapted VGGNet architecture and a residual neural network architecture) for classifying retinal colour fundus images. We trained our model on 76 370 retinal fundus images from 13 099 patients with diabetes who had participated in the Singapore Integrated Diabetic Retinopathy Program, between 2010 and 2013, which has been published previously. In this clinical validation study, we included all patients with a diagnosis of diabetes that attended a mobile screening unit in five urban centres in the Copperbelt province of Zambia from Feb 1 to June 31, 2012. In our model, referable diabetic retinopathy was defined as moderate non-proliferative diabetic retinopathy or worse, diabetic macular oedema, and ungradable images. Vision-threatening diabetic retinopathy comprised severe non-proliferative and proliferative diabetic retinopathy. We calculated the area under the curve (AUC), sensitivity, and specificity for referable diabetic retinopathy, and sensitivities of vision-threatening diabetic retinopathy and diabetic macular oedema compared with the grading by retinal specialists. We did a multivariate analysis for systemic risk factors and referable diabetic retinopathy between AI and human graders. Findings A total of 4504 retinal fundus images from 3093 eyes of 1574 Zambians with diabetes were prospectively recruited. Referable diabetic retinopathy was found in 697 (22•5%) eyes, vision-threatening diabetic retinopathy in 171 (5•5%) eyes, and diabetic macular oedema in 249 (8•1%) eyes. The AUC of the AI system for referable diabetic retinopathy was 0•973 (95% CI 0•969-0•978), with corresponding sensitivity of 92•25% (90•10-94•12) and specificity of 89•04% (87•85-90•28). Vision-threatening diabetic retinopathy sensitivity was 99•42% (99•15-99•68) and diabetic macular oedema sensitivity was 97•19% (96•61-97•77). The AI model and human graders showed similar outcomes in referable diabetic retinopathy prevalence detection and systemic risk factors associations. Both the AI model and human graders identified longer duration of diabetes, higher level of glycated haemoglobin, and increased systolic blood pressure as risk factors associated with referable diabetic retinopathy.Interpretation An AI system shows clinically acceptable performance in detecting referable diabetic retinopathy, vision-threatening diabetic retinopathy, and diabetic macular oedema in population-based diabetic retinopathy screening. This shows the potential application and adoption of such AI technology in an under-resourced African population to reduce the incidence of preventable blindness, even when the ...

No abstract

Background: Early detection and treatment of eye diseases in children is critical in combating childhood blindness. Innovative community-based strategies such as training of teachers in vision screening need to be developed for effective utilisation of the available human resources as well as to counter the challenges of inequitable distribution of trained eye health human resources as well as the limited access of quality eye health care services to the majority of our population. Aim: To evaluate the effectiveness of using teachers as the first level of vision screeners. Materials and Methods: Teacher training programmes were conducted for schoolteachers to educate them about childhood eye diseases and the significance of their early detection. The teachers trained for the school vision screening were from all government, private and community schools located in Kafue District. The teachers then conducted vision screening of learners in their schools. Subsequently, the mobile eye health teams visited the schools for the re-evaluation of learners identified with poor vision. All learners identified with refractive errors had refraction performed on them and spectacles prescribed. The mobile eye health teams referred learners requiring a further ophthalmic evaluation to the University Teaching Hospitals – Eye Hospital which was the base hospital for the programme. The assessment included calculation of true positives, false positives, true negatives and false negatives. Results: One hundred and fifty-four (154) teachers from 73 primary and secondary schools underwent training in vision screening. The teachers screened 18,713 learners and reported eye diseases in 2,818 (15.1%) children. However, the mobile eye health teams examined 5,958 learners who included 2,818 referrals from teachers and 3,140 rescreened learners. The mobile eye health teams confirmed eye problems in 2,818 learners screened by the teachers and further diagnosed more eye problems in 999 learners giving a total of 3,817 learners with eye problems. Thus, the teachers were able to correctly identify eye problems (true positives) in 100.0% (2,818/2,818) of learners. The teachers could not identify eye problems in 999 learners giving false negatives were 26.2% (999/3,817). Conclusion: Considering the high true positive value and the comprehensive coverage provided by the survey, vision screening in schools by teachers is an effective method of identifying learners with eye problems and poor vision early. This strategy could be valuable in reducing the workload of the eye health care staff.