摘要
Abstract
Objective To automatically classify and intelligently identify myopic macular lesions based on ultra-wide-angle fundus imaging technology,and to provide a scientific basis for clinical screening and diagnosis.Methods A retrospective case series study.A total of 1 047 fundus images were selected from 586 patients with myopic maculopathy who were treated in the ophthalmology clinic of our hospital from January 2019 to January 2022,which were marked by two ophthalmologists.The lesion automatic classification model was constructed by residual network,and the focus intelligent recognition model was constructed by DeepLabv3+network.The accuracy,square weighted coefficient k evaluation model and manual diagnosis results are consistent.The accuracy,recall rate and F1 value were used to evaluate the ability of the model to identify lesions,and the area under the receiver operating characteristic(ROC)curve sensitivity,specificity and consistency index were used to evaluate the value of the model in the diagnosis of pathological myopia.Results The classi-fication accuracy of the automatic classification model of myopic maculopathy was 0.9304(95%CI:0.8619～0.9761)and the square weighting coefficient k was 0.9392.The accuracy of the intelligent lesion recognition and segmentation model was 0.9215(95%CI:0.8799～0.9613),and the square weighting coefficient k was 0.9165.The image level F1 values of the segmentation model in identifying optic disc,para-optic disc atrophy,diffuse atrophy,patchy atrophy and macular at-rophy were all more than 0.90.The area under the receiver operating characteristic curve of the common decision model for the diagnosis of pathological myopia was 0.9890(95%CI:0.8154～0.9978),the sensitivity was 0.9671(95%CI:0.8513～0.9758),the specificity was 0.9905(95%CI:0.7627～0.9989),and the consistency index was 0.9827(95%CI:0.7362～0.9931).Conclusions Based on the ultra-wide-angle fundus imaging technology,the accuracy of the automatic classifica-tion and intelligent recognition model of myopic maculopathy is high,which is beneficial to the prevention,control,screening and diagnosis of myopia and improve the work efficiency.关键词
超广角眼底成像技术/病理性近视/近视性黄斑病变/病灶识别Key words
Ultra-wide-angle fundus imaging/Pathological myopia/Myopic maculopathy/Focus recognition
