Artificial intelligence using deep learning to screen for referable and vision-threatening diabetic retinopathy in Africa: a clinical validation study
Artificial intelligence using deep learning to screen for referable and vision-threatening diabetic retinopathy in Africa: a clinical validation study
复制标题
DOI：
10.1016/s2589-7500(19)30004-4
复制
发表时间：
2019-05-01
影响因子：
30.8
通讯作者：
Ting, Daniel S. W.
中科院分区：
文献类型：
--
作者：
Bellemo, Valentina;Lim, Zhan W.;Ting, Daniel S. W.
背景 需要采取激进措施来识别和减少糖尿病引起的失明，以实现到 2030 年的可持续发展目标。因此，我们在中低收入国家赞比亚的基于人群的糖尿病视网膜病变筛查项目中使用深度学习评估了人工智能 (AI) 模型的准确性。方法我们采用了由两个卷积神经网络（改编的 VGGNet 架构和残差神经网络架构）组合组成的整体 AI 模型，用于对视网膜彩色眼底图像进行分类。我们使用 2010 年至 2013 年间参加过新加坡综合糖尿病视网膜病变项目的 13 099 名糖尿病患者的 76 370 幅视网膜眼底图像来训练我们的模型，该项目之前已发表。在这项临床验证研究中，我们纳入了2012年2月1日至6月31日期间在赞比亚铜带省五个城市中心参加流动筛查的所有诊断为糖尿病的患者。在我们的模型中，可转诊的糖尿病视网膜病变被定义为中度非增殖性糖尿病视网膜病变或更严重、糖尿病黄斑水肿和无法分级的图像。威胁视力的糖尿病视网膜病变包括严重的非增殖性和增殖性糖尿病视网膜病变。与视网膜专家的分级相比，我们计算了可转诊的糖尿病视网膜病变的曲线下面积 (AUC)、敏感性和特异性，以及威胁视力的糖尿病视网膜病变和糖尿病黄斑水肿的敏感性。我们对 AI 和人类评分者之间的系统性危险因素和可参考的糖尿病视网膜病变进行了多变量分析。 结果 前瞻性招募了来自 1574 名赞比亚糖尿病患者的 3093 只眼睛的总共 4504 张视网膜眼底图像。 697 只眼 (22.5%) 发现需转诊的糖尿病视网膜病变，171 只眼 (5.5%) 发现威胁视力的糖尿病视网膜病变，249 只眼 (8.1%) 发现糖尿病黄斑水肿。 AI系统针对可转诊糖尿病视网膜病变的AUC为0.973（95% CI 0.969-0.978），相应的敏感性为92.25%（90.10-94.12），特异性为89.04%（87.85-90.28）。威胁视力的糖尿病视网膜病变敏感性为99.42%（99.15-99.68），糖尿病黄斑水肿敏感性为97.19%（96.61-97.77）。人工智能模型和人类评分者在可参考的糖尿病视网膜病变患病率检测和系统性危险因素关联方面显示出相似的结果。 AI 模型和人类评分者都将较长的糖尿病病程、较高的糖化血红蛋白水平和较高的收缩压确定为与可转诊的糖尿病视网膜病变相关的危险因素。 解释 在基于人群的糖尿病视网膜病变筛查中，AI 系统在检测可转诊的糖尿病视网膜病变、威胁视力的糖尿病视网膜病变和糖尿病黄斑水肿方面显示出临床上可接受的性能。这表明，即使模型是在不同人群中进行训练的，这种人工智能技术在资源贫乏的非洲人口中的潜在应用和采用也可以减少可预防性失明的发生率。版权所有 (C) 2019 作者。由爱思唯尔有限公司出版
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 model is trained in a different population. Copyright (C) 2019 The Author(s). Published by Elsevier Ltd.