He Xingru, Dai Guangzheng, Che Huixin, Zhang Chenguang, Yan Hairu, Dang Yu, Dong Haifeng
School of Public Health, He University, Shenyang, Liaoning, China.
Department of Clinical Research Center, He Eye Specialist Hospital, Shenyang, Liaoning, China.
NPJ Digit Med. 2025 Jul 17;8(1):449. doi: 10.1038/s41746-025-01853-2.
Primary angle closure glaucoma (PACG) is a major cause of irreversible blindness, characterized by shallow anterior chambers and narrow angles. In our study, we used anterior segment photographs from a modified mobile camera to develop algorithms for identifying narrow anterior chamber angles (NACA). We calculated eight biological parameters after preprocessing the images, segmenting the corneal and iris light bands, and modeling the central anterior chamber. In the training dataset, the accuracy of NACA identification using these parameters ranged from 0.68 to 0.85, with an AUC of 0.76 to 0.90. The internal test dataset's accuracy ranged from 0.65 to 0.85, with sensitivity between 0.40 and 0.95 and specificity between 0.61 and 1.00. An ensemble model achieved a sensitivity of 0.90 and a specificity of 0.85 on the internal test dataset, but its performance declined on external datasets. Despite generalization challenges, portable slit lamps equipped with advanced algorithms show promise for NACA screening.
原发性闭角型青光眼（PACG）是不可逆性失明的主要原因，其特征为前房浅和房角狭窄。在我们的研究中，我们使用来自改良移动相机的眼前节照片开发了用于识别窄前房角（NACA）的算法。我们在对图像进行预处理、分割角膜和虹膜光带以及对中央前房进行建模后计算了八个生物学参数。在训练数据集中，使用这些参数进行NACA识别的准确率在0.68至0.85之间，曲线下面积（AUC）在0.76至0.90之间。内部测试数据集的准确率在0.65至0.85之间，灵敏度在0.40至0.95之间，特异性在0.61至1.00之间。一个集成模型在内部测试数据集上实现了0.90的灵敏度和0.85的特异性，但其在外部数据集上的性能有所下降。尽管存在泛化挑战，但配备先进算法的便携式裂隙灯在NACA筛查方面显示出前景。
