摘要
目的:调查重庆市18岁及以上人群高度近视的患病率及影响因素、高度近视视网膜脉络膜病变患病率及视力损伤情况;并比较人工智能(AI)软件与人工阅片诊断高度近视视网膜脉络膜病变的一致性。方法:横断面调查研究。采用分层整群抽样方法于2021年7月对重庆市永川区18岁及以上人群进行眼健康调查,内容包括问卷调查和视力、等效球镜度(SE)、非接触眼压、裂隙灯显微镜和眼底照相等眼科检查。高度近视诊断标准为SE≤-6.00 D,如伴后巩膜葡萄肿、Fuchs斑、漆裂纹、豹纹状改变及视盘周萎缩弧等眼底退行性改变诊断为高度近视视网膜脉络膜病变。筛选SE≤-6.00 D眼底图,通过AI及人工阅片方式分析,诊断高度近视视网膜脉络膜病变。采用卡方检验和Logistic回归分析对高度近视患病的影响因素进行分析;采用Kappa检验和受试者特征(ROC)曲线分析比较人工及AI阅片的一致性。结果:方案设定流调对象4364人,实际调查4695人,应答率为107.58%。SE≤-6.00 D为237人(384眼),重庆市高度近视患病率5.05%;相对于18~30岁的人群,60~70岁年龄段患高度近视风险降低(OR=0.38,95%CI:0.188~0.760,P=0.006);女性患高度近视风险增加(OR=1.41,95%CI:1.028~1.935,P=0.033);高中、大专、大学及以上人群较小学及以下人群更容易患高度近视(OR=3.62、3.34、7.30,P<0.001)。人工阅片诊断高度近视视网膜脉络膜病变162人,占高度近视人群的68.35%,总人群患病率为3.45%;AI诊断高度近视视网膜脉络膜病变改变者155人,占高度近视人群的65.40%,总人群患病率为3.30%;人工阅片与AI诊断高度近视视网膜脉络膜病变(Kappa=0.895,P=0.001)、同时诊断病理性近视(Kappa=0.931,P=0.001),表明AI与人工阅片诊断一致性良好。AI诊断高度近视视网膜脉络膜病变ROC曲线下面积(AUC=0.959,P<0.001)、病理性近视(AUC=0.955,P<0.001),灵敏度和特异度均在90%以上,表明AI诊断准确性较高。同时统计高度近视视网膜脉络膜病变患者视力,其中视力损伤28人,盲13人,人群致盲率为0.28%。结论:重庆市18岁及以上人群高度近视患病率较高,呈低龄化趋势;年龄、性别、教育程度是高度近视患病的影响因素。AI与人工阅片在高度近视视网膜脉络膜病变诊断方面一致性较高。
Objective:To investigate the prevalence of high myopia and its influencing factors,the prevalence of high myopic choroidal retinopathy and visual impairment in people aged 18 and above in Chongqing;and compare the consistency between artificial intelligence(AI)software and manual film reading in the diagnosis of high myopic choroidal retinopathy.Methods:This was a cross-sectional study.The stratified cluster sampling method was adopted to investigate the eye health of the people aged 18 years and above in Yongchuan District,Chongqing in July 2021,including questionnaire survey and eye examinations including visual acuity,equivalent spherical lens(SE),non-contact intraocular pressure,slit lamp,and fundus photography.High myopia was defined as SE≤-6.00 D and had high myopic choroidal retinopathy,such as posterior staphyloma,Fuchs spot,lacquer crack,retina leopard pattern change,and optic disc atrophy arc,etc.The fundus images of people with SE≤-6.00 D were screened and analyzed by AI and manual reading methods to determine high myopia retinochoroidal disease.Chi-square test and logistic regression analysis were used to analyze the influencing factors of high myopia.Kappa test and receiver operating characteristic(ROC)curve analysis were used to compare the consistency of manual and AI film reading.Results:A total of 4364 people were surveyed in the scheme,4695 were actually investigated,with the response rate was 107.58%,including 237 people with 384 eyes SE≤-6.00 D,the prevalence of high myopia in Chongqing was 5.05%.People between the ages of 18 and 30 were more likely to have high myopia(OR=0.38,95%CI:0.188-0.760,P=0.006);Women had a higher risk of high myopia(OR=1.41,95%CI:1.028-1.935,P=0.033);People in high school,college,university or above were more likely to suffer from high myopia than those in primary school or below(OR=3.62,3.34,7.30,P<0.001).One hundred and sixty two people diagnosed with high myopic choroidal retinopathy by manual reading,accounting for 68.35%of high myopia population,and the overall prevalence rate was 3.45%;155 people diagnosed with high myopia retinochoroidal disease changes by AI,accounting 65.40%,and the overall prevalence rate was 3.30%.Artificial photoreading and AI were used to diagnose retinal choroidopathy in high myopic choroidal retinopathy(Kappa=0.895,P<0.001)and pathological high myopia(Kappa=0.931,P<0.001),indicating that AI and manual were in good agreement in diagnosis.The area under ROC curve of high myopia(AUC=0.959,P<0.001)and pathological myopia(AUC=0.955,P<0.001)were diagnosed by AI in the ROC test,and the sensitivity and specificity were both above 90%,indicating that AI diagnosis had high diagnostic accuracy.At the same time,the visual acuity of patients with high myopic choroidal retinopathy was analyzed,among them,28 patients had visual impairment and 13 were blind,and the population blindness rate was 0.28%.Conclusions:The prevalence of high myopia in people aged 18 and above in Chongqing is relatively high,and presents a younger trend.Age,gender,and education are the risk factors for high myopia.The diagnosis in high myopic choroidal retinopathy was consistent between AI and manual reading.
作者
彭镜
姚浩
李兰娇
张聪
周希瑗
Jing Peng;HaoYao;Lanjiao Li;Cong Zhang;Xiyuan Zhou(Chongqing Key Laboratory of Ophthalmology,the Second Affiliated Hospital of Chongqing Medical University,Chongqing 400010,China)
出处
《中华眼视光学与视觉科学杂志》
CAS
CSCD
2023年第10期739-747,共9页
Chinese Journal Of Optometry Ophthalmology And Visual Science
基金
国家自然科学基金(82070976)
