| Background:Diabetic retinopathy（DR）is one of the main chronic complications of diabetes and one of the main causes of blindness in adults[1–3],It has attracted more and more attention from endocrinologists and ophthalmologists.Studies have shown that DR Depending on the country,region,and ethnicity,the prevalence of developing countries is lower than that of developed countries[4].A meta-analysis included22,896 diabetic patients from 35 studies around the world[5].The results of the study showed that the prevalence of DR was 34.6%,of which proliferative diabeticretinopathy（PDR）is 6.96%,diabetic macular edema（DME）is 6.81%,and vision-threatening DR is 10.2%.According to relevant research results,mainland China The prevalence of DR in the diabetic population was 23%（95%CI:17.8%-29.2%）,of which the proportion of PDR（proliferative diabetic retinopathy,PDR）was 2.8%（1.9%-4.2%）,and the proportion of non-PDR was 19.1%（13.6%～26.3%）,the rural area is higher than the city,the north is higher than the south and the east[6–7].The prevalence of DR among Chinese in Singapore is 20.1%[8],among Chinese in the United States,the prevalence of DR among Chinese in the United States is DR.It is 25.7%[9].The prevalence of DR in Taiwan is as high as 35%[10].The high incidence of diabetic retinopathy has brought huge pain and economic burden to patients.According to the data of the sixth large-scale epidemiological survey report on diabetic patients in China,the total number of diabetic patients in mainland China is about 129 million.The diagnosis and control of diabetes in this epidemiological survey are based on ADA and WHO standards.The basis is that Hb A1c is the standard.The prevalence of diabetes in China is the same as in other countries in the world.The total prevalence of diabetes diagnosed according to WHO standards has increased from 9.7%in 2007 and 2010 to 10.4%in 2013.By 2017,11.2%[11]in this study,the prevalence rate has been increasing year by year.The total prevalence of diabetes in Chinese adults is 12.8%（12.0%-13.6%）,and the prevalence of self-reported diabetes is 6.0%（5.4%～6.7%）,the total standardized prevalence rate of newly diagnosed diabetes using ADA standards is 6.8%（6.1%to 7.4%）,and the total prevalence of diabetes in men is higher than that in women.According to the diagnostic criteria of ADA,the current adult diabetes in China The prevalence in the early stage is about 35.2%（33.5%～37.0%）.Patients with type 2 diabetes may have different degrees of diabetic retinopathy at the time of diagnosis,and the severity of diabetic retinopathy is directly related to the severity of the diabetic patient and the length of the disease,which can lead to severe damage to visual function.The earlier the screening of diabetic retinal disease,the more beneficial it is to control the degree of disease,but the screening for retinal disease is faced with the following:（1）There are many patients and few doctorsAt present,the number of diabetic patients in China is about 129.8 million,and the DR lesion rate of diabetic patients accounts for about 23%,that is,about 30million people have diabetic fundus retinopathy.According to the National Health Commission issued in June 2020,According to the survey data of the White Paper on China’s Eye Health[12],there are currently 44,800 ophthalmologists in the country,a significant increase from 19,000 in 2003;however,optometrists are still extremely scarce,increasing from more than 1,400 in 2003 to There are more than 6,000ophthalmologists,and the gap is still very large.There are only 1.6 ophthalmologists per 50,000 people,which is just an average situation.In some areas,there may not even be 0.6 ophthalmologists out of 50,000 people,which means that when a patient happens retinal fundus diseases cannot be diagnosed and screened in time,and they cannot be controlled and treated in time.（2）The screening method is not convenient enoughIt is understood that diabetes is a metabolic disease characterized primarily by high blood sugar levels,which are caused by inadequate insulin secretion or impaired biological function,or both,which are themselves endocrine disease,fundus retinopathy is a kind of ocular disease with specific changes,so it needs the help of endocrinologists and ophthalmologists in the diagnosis and treatment of fundus diseases.Even general practitioners can not be fully equipped with the ability to accurately screen for lesions,which adds a huge workload to doctors in our country,where doctors are not high per capita,and for patients in remote areas,it is impossible to detect the disease in time.When the disease develops to the later stage,the patient’s visual function will be damaged irreversibly,and the possibility of vision loss or even blindness will occur even if the disease can be screened out at the later stage,also can not obtain the effective treatment from the root.So it’s important to find a faster and more convenient way to screen for diabetic retinopathy when the ratio of doctors to patients is out of whack.Since diabetic retinopathy（DR）is the primary cause of blindness in DM,early screening for DR has important clinical significance.The screening of DR lesions is mainly diagnosed by fundus photography.How to accurately and quickly analyze fundus photos has become an important issue.An important challenge.Artificial intelligence is a new technology that simulates and expands human intelligence through research and development.In recent years,with the development of the Internet and computer software and hardware,the development of artificial intelligence in deep learning has been greatly improved.,Forming research directions such as intelligent robots,speech recognition,pattern recognition,image recognition,expert systems,natural language processing,etc.,making artificial intelligence also begin to have different degrees of application in different fields such as industry,medical care,finance,and security,such as humans.Face recognition and surgical robots have brought great convenience to people’s work and life.This article mainly introduces the application of artificial intelligence in the analysis of fundus photographic images of DR lesions.Purpose:Through the artificial analysis of the fundus photography of diabetic patients and the consistency of the results of the analysis of the fundus photos by artificial intelligence analysis,the reliability of artificial intelligence screening is verified.The application of artificial intelligence analysis will be explored for the significance and value of future fundus screening.Method:This study adopted a retrospective research method and included 1053 cases of2106 eyes of diabetic patients who were continuously admitted to the Endocrinology Department of the First Affiliated Hospital of Zhengzhou University from May 2018to May 2019.Among them,888 were males and 165 were females;age 20 to 70 years old,with an average age of 53 years.All patients used the Japanese Kowa non-mydriatic fundus camera to perform fundus examinations on diabetic patients.The artificial intelligence analysis of the Shanggong ophthalmology cloud network screening platform was used to automatically detect exudation,bleeding,and microaneurysms Wait for the characteristic lesions of diabetic retinopathy（DR）and identify them,and at the same time,automatically classify the image detection results according to the DR international staging standard.Manual analysis is analyzed by two attending physicians in endocrinology and reviewed by the chief physician to ensur the accuracy of manual analysis.When there is a difference between the analysis results of the two analysis methods,the manual analysis result is the standard.Calculate and compare the agreement rate of the two analysis methods.The agreement rate=（the same number of eyes with the same diagnosis result/the total number of effective eyes collected）×100%.Kappa consistency test is performed on the results of manual analysis and artificial intelligence analysis,0≤κ<0.2 is very poor consistency,0.2≤κ<0.4 is poor consistency,0.4≤κ<0.6 is medium consistency,0.6≤κ<1 is better consistency.Results:Of the 2106 eyes,64 eyes that could not be intelligently identified due to severe disease were excluded,and 2042 eyes were finally included in the analysis.1835 eyes were completely consistent with artificial intelligence analysis results,accounting for89.86%;there were 207 eyes with differences in the analysis,accounting for 10.14%.The difference between the two is mainly manifested as:（1）artificial intelligence analyzes spotting bleeding and exudation,while manual analysis of 96 normal eyes（96/2042,4.70%）;（2）artificial intelligence analysis of drusen,Artificial analysis showed that 71 eyes were spotted exudation（71/2042,3.48%）;（3）Artificial intelligence analyzed normal or vitreous degeneration,while artificial analysis spotted exudation or bleeding or microaneurysms in 40 eyes（40/2042,1.95%）.The diagnostic rates of manual analysis and artificial intelligence analysis for DR were23.2%and 20.2%,respectively,and the diagnostic rates for non-DR were 76.8%and79.8%.The Kappa consistency test results showed that manual analysis and manual analysis The diagnosis results of intelligent analysis showed moderate consistency（κ=0.576,P<0.01）.Conclusion:Artificial analysis and artificial intelligence analysis showed moderate consistency in the diagnosis of fundus lesions in diabetic patients.The accuracy of artificial intelligence interpretation reached 87.7%,and the misdiagnosis rate was12.2%.In summary,artificial intelligence analysis is used as an early screening method for fundus lesions.Yes,but it needs further improvement.For severe fundus lesions,artificial intelligence analysis still relies on human interpretation.In the future,with the development of technology,artificial intelligence will definitely play a greater role. |
PDF Full Text Request