Xie He, Li Zhongwen, Wu Chengchao, Zhao Yitian, Lin Chengmin, Wang Zhouqian, Wang Chenxi, Gu Qinyi, Wang Minye, Zheng Qinxiang, Jiang Jiewei, Chen Wei
National Clinical Research Center for Ocular Diseases, Eye Hospital, Wenzhou Medical University, Wenzhou, China.
Ningbo Eye Hospital, Wenzhou Medical University, Ningbo, China.
Front Cell Dev Biol. 2023 Jul 28;11:1197239. doi: 10.3389/fcell.2023.1197239. eCollection 2023.
To develop a visual function-based deep learning system (DLS) using fundus images to screen for visually impaired cataracts. A total of 8,395 fundus images (5,245 subjects) with corresponding visual function parameters collected from three clinical centers were used to develop and evaluate a DLS for classifying non-cataracts, mild cataracts, and visually impaired cataracts. Three deep learning algorithms (DenseNet121, Inception V3, and ResNet50) were leveraged to train models to obtain the best one for the system. The performance of the system was evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. The AUC of the best algorithm (DenseNet121) on the internal test dataset and the two external test datasets were 0.998 (95% CI, 0.996-0.999) to 0.999 (95% CI, 0.998-1.000),0.938 (95% CI, 0.924-0.951) to 0.966 (95% CI, 0.946-0.983) and 0.937 (95% CI, 0.918-0.953) to 0.977 (95% CI, 0.962-0.989), respectively. In the comparison between the system and cataract specialists, better performance was observed in the system for detecting visually impaired cataracts ( < 0.05). Our study shows the potential of a function-focused screening tool to identify visually impaired cataracts from fundus images, enabling timely patient referral to tertiary eye hospitals.
开发一种基于视觉功能的深度学习系统（DLS），利用眼底图像筛查视力受损的白内障。从三个临床中心收集了总共8395张眼底图像（5245名受试者）以及相应的视觉功能参数，用于开发和评估一个用于对非白内障、轻度白内障和视力受损白内障进行分类的DLS。利用三种深度学习算法（DenseNet121、Inception V3和ResNet50）训练模型，以获得系统的最佳算法。使用受试者工作特征曲线下面积（AUC）、灵敏度和特异性评估系统的性能。最佳算法（DenseNet121）在内部测试数据集和两个外部测试数据集上的AUC分别为0.998（95%CI，0.996 - 0.999）至0.999（95%CI，0.998 - 1.000）、0.938（95%CI，0.924 - 0.951）至0.966（95%CI，0.946 - 0.983）以及0.937（95%CI，0.918 - 0.953）至0.977（95%CI，0.962 - 0.989）。在系统与白内障专科医生的比较中，系统在检测视力受损白内障方面表现更优（<0.05）。我们的研究表明，一种以功能为重点的筛查工具具有从眼底图像中识别视力受损白内障的潜力，能够及时将患者转诊至三级眼科医院。
