From the Department of Electrical and Computer Engineering (M.V.), Isfahan University of Technology, Isfahan, Iran.
Farabi Eye Hospital (M.M., N.Z., M.S.M.A.F.), Tehran University of Medical Sciences, Tehran, Iran.
Am J Ophthalmol. 2023 Aug;252:1-8. doi: 10.1016/j.ajo.2023.02.016. Epub 2023 Mar 2.
A deep learning framework to differentiate glaucomatous optic disc changes due to glaucomatous optic neuropathy (GON) from non-glaucomatous optic disc changes due to non-glaucomatous optic neuropathies (NGONs).
Cross-sectional study.
A deep-learning system was trained, validated, and externally tested to classify optic discs as normal, GON, or NGON, using 2183 digital color fundus photographs. A Single-Center data set of 1822 images (660 images of NGON, 676 images of GON, and 486 images of normal optic discs) was used for training and validation, whereas 361 photographs from 4 different data sets were used for external testing. Our algorithm removed the redundant information from the images using an optic disc segmentation (OD-SEG) network, after which we performed transfer learning with various pre-trained networks. Finally, we calculated sensitivity, specificity, F1-score, and precision to show the performance of the discrimination network in the validation and independent external data set.
For classification, the algorithm with the best performance for the Single-Center data set was DenseNet121, with a sensitivity of 95.36%, precision of 95.35%, specificity of 92.19%, and F1 score of 95.40%. For the external validation data, the sensitivity and specificity of our network for differentiating GON from NGON were 85.53% and 89.02%, respectively. The glaucoma specialist who diagnosed those cases in masked fashion had a sensitivity of 71.05% and a specificity of 82.21%.
The proposed algorithm for the differentiation of GON from NGON yields results that have a higher sensitivity than those of a glaucoma specialist, and its application for unseen data thus is extremely promising.
建立一种深度学习框架，以区分由青光眼视神经病变（GON）引起的青光眼性视盘改变与由非青光眼视神经病变（NGON）引起的非青光眼性视盘改变。
横断面研究。
使用 2183 张数字眼底彩色照片，训练、验证和外部测试深度学习系统，以将视盘分类为正常、GON 或 NGON。采用单中心数据集（1822 张图像，NGON 图像 660 张，GON 图像 676 张，正常视盘图像 486 张）进行训练和验证，而来自 4 个不同数据集的 361 张照片用于外部测试。我们的算法使用视盘分割（OD-SEG）网络从图像中去除冗余信息，然后使用各种预训练网络进行迁移学习。最后，我们计算敏感性、特异性、F1 评分和精度，以显示判别网络在验证和独立外部数据集的性能。
对于分类，单中心数据集性能最佳的算法是 DenseNet121，其敏感性为 95.36%，特异性为 92.19%，准确性为 95.35%，F1 得分为 95.40%。对于外部验证数据，我们的网络区分 GON 和 NGON 的敏感性和特异性分别为 85.53%和 89.02%。以掩蔽方式诊断这些病例的青光眼专家的敏感性为 71.05%，特异性为 82.21%。
本研究提出的用于区分 GON 和 NGON 的算法具有比青光眼专家更高的敏感性，其对未见数据的应用极具潜力。
