Overcome the barriers of vision care in Africa using smartphones

项目总结/摘要 全世界有6.71亿人患有未矫正的屈光不正，其中1.61亿人患有距离 由于未矫正的屈光不正（URE）造成的视力损害。URE导致的远视力损害 特别影响到儿童和年轻人，限制了他们的教育活动和社会生活。虽然 屈光不正的患病率较低，非洲因URE而失明的患病率是非洲的6倍， 世界高收入地区。这表明非洲在获得眼科护理方面存在广泛的差距， 导致长期的负面影响。事实上，每百万人中只有2.5名眼科医生 在撒哈拉以南非洲地区，每百万人中约有32名眼科医生。创新 需要采取以社区为基础的办法，以解决提供所需保健的日益沉重的负担。 为了响应赠款机会，以改善低收入和中等收入国家的医疗保健服务， 移动的健康，我们建议开发一个AI驱动的移动的应用程序，用于测量屈光不正 并通过在埃塞俄比亚进行实地测试来验证它。凭借其相对简单的协议，该应用程序可以允许 人们仅使用智能手机来测量近视屈光不正，而不需要任何电话附件。 因此，它克服了成本和可获得性的主要障碍。最近一项关于这款折射应用的研究显示， 与美国和中国多中心研究中的屈光临床测量结果具有良好的一致性。给定 在R21阶段，人工智能算法在代表性不足的非洲种族中的偏见和不准确性， 我们将在来自非洲的人脸数据集上训练一个自定义神经网络，以制作当前版本的应用程序。 更适合黑皮肤的非洲人我们将验证应用程序在埃塞俄比亚的折射测量， 将其与标准临床测量进行比较，并通过确定视觉上的改善， 由基于app的屈光矫正产生的敏锐度。在验证了用于以下目的的移动的技术之后， 处方负担得起的眼镜在医院设置，我们将进一步评估社区视力保健计划 在R33阶段，包括视力测试和眼镜分发。社区志愿者将接受培训 使用该应用程序进行视力筛查，并将在埃塞俄比亚农村进行实地测试，以确定 以社区为基础的近视筛查方法在现实世界中的敏感性和特异性。与 移动健康解决方案，我们准备对服务不足的非洲的眼科护理产生重大影响。

PROJECT SUMMARY / ABSTRACT Worldwide, 671 million people have uncorrected refractive error out of which 161 million people have distance vision impairments due to uncorrected refractive error (URE). Distance vision impairment due to URE especially affects children and young adults, limiting their educational activities and social life. Although the prevalence of refractive error is lower, the prevalence of blindness due to URE in Africa is 6 times as that in high-income regions of the world. This indicates widespread disparities in access to eye care in Africa, which leads to long term negative effects. As a matter of fact, there are only 2.5 ophthalmologists per million population in Sub-Saharan Africa, against a global mean of about 32 ophthalmologists per million. Innovative community-based approaches are needed to tackle the increasing burden of providing the needed healthcare. In response to the grant opportunity to improve healthcare access to low and middle-income countries using mobile health, we propose to develop an AI-powered mobile application for measurement of refractive error and validate it by performing field tests in Ethiopia. With its relatively simple protocol, the app can allow lay people measure myopic refractive error using only a smartphone, without needing any phone attachments. Thus, it overcomes the main barriers of cost and accessibility. A recent study on this refraction app showed good consistency with clinical measurements of refraction in a multi-center study in the US and China. Given the bias and inaccuracies of AI algorithms in people of underrepresented African ethnicities, in the R21 phase, we will train a custom neural network on face datasets of African origin to make the current version of the app more suitable for dark-skinned Africans. We will validate the app measurements of refraction in Ethiopia by comparing it with standard clinical measurement, and also by determining the improvements in the visual acuity resulting from the app-based refractive correction. After the validation of the mobile technology for prescribing affordable eyeglasses in hospital settings, we will further evaluate community vision care programs in the R33 phase, including vision testing and dispensing of eyeglasses. Community volunteers will be trained to perform vision screening with the app and a field test will be performed in rural Ethiopia to determine the sensitivity and specificity of community-based myopia screening approach in real-world settings. With the mHealth solution, we are poised to make significant impact on the eye care in under-served Africa.