Artificial intelligence using deep learning to screen for referable and vision-threatening diabetic retinopathy in Africa: a clinical validation study

Summary 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 retinopathyscreening 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 networkarchitecture) 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 RetinopathyProgram, 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 retinopathywas defined as moderate non-proliferative diabetic retinopathyor worse, diabetic macular oedema, and ungradable images. Vision-threatening diabetic retinopathycomprised 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 retinopathyand diabetic macular oedemacompared with the grading by retinal specialists. We did a multivariate analysis for systemic riskfactors and referable diabetic retinopathybetween 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 retinopathywas found in 697 (22·5%) eyes, vision-threatening diabetic retinopathyin 171 (5·5%) eyes, and diabetic macular oedemain 249 (8·1%) eyes. The AUC of the AI system for referable diabetic retinopathywas 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 retinopathysensitivity was 99·42% (99·15–99·68) and diabetic macular oedemasensitivity was 97·19% (96·61–97·77). The AI model and human graders showed similar outcomes in referable diabetic retinopathyprevalence detection and systemic riskfactors 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 oedemain population-based diabetic retinopathyscreening. 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 model is trained in a different population. Funding National Medical Research Council Health Service ResearchGrant, Large Collaborative Grant, Ministry of Health, Singapore; the SingHealth Foundation; and the Tanoto Foundation.