There are about 110 million diabetic patients in China, and diabetic retinopathy is one of the most serious complications. The incidence rate is 24.7%-37.5%. With the advent of an aging society, the incidence of age-related macular degeneration continues to rise. However, at the current stage, the shortage of professional ophthalmologists in China, lack of medical resources in primary hospitals, and inefficient diagnostic methods have always restricted the diagnosis of ophthalmology. The long-range vision intelligent viewing platform created by the combination of artificial intelligence (AI) and ophthalmology can accurately diagnose, quantify, and timely screen the patient's fundus image through intelligent reading, which will effectively assist doctors in diagnosis and improve diagnosis efficiency. Especially for regions where doctors are relatively scarce, through efficient telemedicine systems, primary hospitals are expected to gradually approach the same level of diagnosis as large hospitals. Ophthalmic grading treatment becomes possible Nowadays, the form of eye health has become increasingly severe, and young people have high myopia. According to data released by the National Health and Family Planning Commission, more than 50% of people did not receive regular eye examinations, and more than 90% of the patients were treated after onset. Actually, early intervention can be made for elderly eye diseases and chronic diseases. Auror Ophthalmic's subsidiary Hunan Vision Technology Co., Ltd., as the main body of its Internet development strategy, has been serving eye health examinations, medical optometry, eye health products, personal eye health doctors, remote diagnosis and treatment services, and referral services. It has created an online and offline eye health "Internet +" service model. However, the current status of China's ophthalmological medical service industry is “decentralized and regionally concentratedâ€. The development of technologies such as telemedicine, intelligent equipment, and big data also allows graded diagnosis and treatment to be gradually implemented in ophthalmology. Intel Joint Eyes, Extreme Vision, Jinhong Technology Introduce Artificial Intelligence Disease Recognition Solution for Eyes, Build a Long-range Remote Intelligent Reading Platform, and Further Improve Diabetic Retinopathy (DR) and Age-related Macular Degeneration (AMD) Screening Accuracy and efficiency. Neighboring Remote Intelligent Reading Platform “The long-distance vision reading platform on the one hand realizes grading diagnosis and treatment on the one hand, and realizes eye health management on the other.†said Xu Ming, general manager of Hunan AELINK Internet Technology Co., Ltd., doctors at the basic level are very lacking, and the reading platform can be centralized. Resources to address the fragmented demand for decentralization. Collect data from community health centers, optometry centers, county-level hospitals, optical shops, and other institutions, perform intelligent readings on the remote intelligent interpretation platform, and determine that abnormal films are sent to the film reader for interpretation. Finally, all Fundus photographs are used to accumulate data and continue learning. The diagnosis results are fed back to the patient and a referral is made. “Obsteative remote intelligent interpretation platform will play a role at the grass-roots level because of the scarcity of primary-level ophthalmologists and insufficient diagnostic capabilities.†Sun Tao, deputy general manager and CTO of Hunan Aurora Connect Technology Co., Ltd. said that by investing equipment in community health services In the center, the fundus photographs can be initially screened by machine autonomous learning, and photographs in question can be sent back to the interpreting doctor for a second judgment. At present, nearly 20 community service centers and more than 10 Aier Ophthalmology self-built community eye health service centers have begun to use remote and intelligent reading platforms. It is expected to reach 50 by the end of the year, and the major pilot cities will be located in Changsha and Beijing. Chengdu, Tianjin, Shenzhen. AI reading only takes 10 seconds Neighboring remote intelligent reading platform is divided into software and hardware. The software is artificial intelligence + reading system, hardware is the reading center + doctor team. Neighboring remote intelligent interpretation platform uses Convolutional Neural Network (CNN). CNN belongs to a branch of machine learning and realizes pattern recognition. Aiming at the remote intelligent reading platform, DR and AMD were first selected as the artificial intelligence screening disease because the two diseases had the highest prevalence rate in the ophthalmic industry, and feature extraction was relatively convenient. Sun Tao pointed out that the platform through a 100,000-labeled fundus photo of the fundus for a year and a half of learning, to achieve the recognition accuracy of DR and AMD reached 93%, the identification of other diseases is also gradually research and development. Aier Ophthalmic and Aurora Connect, Intel, and Jinhong Technology each play different roles. Aier Eye provides learning by providing hundreds of hospitals across the country to provide fundus photos as a platform for learning and developing and operating the viewing platform. Intel conducted corresponding research and development of the chip for the interpretation of the film. Jin Hong Technology provided the production of fundus camera equipment and artificial intelligence model. The remote vision reading platform and picture storage are all on the third-party cloud platform. The artificial intelligence operation is on the local data center, and the three are connected via a dedicated network. Zhang Xiaobo, head of the Intel Data Center Medical and Life Sciences Group in Asia Pacific, said that the vision of the remote intelligent interpretation platform is a cloud-to-end practice. Based on the existing AI of Aier Ophthalmology, combined with the features of ophthalmologic image analysis processing, Intel provides customized hardware solutions from the perspective of processor and related hardware optimization, including the framework of artificial intelligence algorithms, Caffe, and overall support optimization. , Help the platform's calculations to improve 3-5 times and more accurate, image judgment time reduced from 1 minute to less than 10 seconds. Due to the different qualifications of the doctor reading the film, some doctors cannot determine the condition, and the photo will be submitted to the next level reader for reading. The platform also developed reading track records so that the primary reader can see the final reading results. Regardless of any fundus camera, the remote intelligent reading platform can analyze the photos, so the platform is very popular in promotion. The most important promotion issue is the deployment of fundus cameras. Zheng Zhuming, chairman and co-founder of Jin Hong Technology (Taiwan) Co., Ltd., said that the application of handheld fundus inspection equipment in the future will greatly accelerate the use of the platform.