[关键词]
[摘要]
青光眼的早期诊断和病情进展的监测是一项非常复杂的工作，需要综合评估视神经的结构(眼底照相或光学相干断层扫描)和功能(视野)损伤，在很大程度上取决于专业医生的临床经验。人工智能(AI)在眼科的应用提高了我们对青光眼的理解，并能帮助减少处理相关临床任务所需的人力和时间。随着深度学习的到来，出现了许多用于眼科图像分类、分割和增强的工具。尤其最近3a，已经提出了多种适用于青光眼的AI方法，通过对功能和/或结构的分析来帮助诊断青光眼，并且探索了使用AI来监测疾病进展的方法，提高了判断疾病预后的可靠性，给个体化精准医疗带来可能。然而，这些算法还有待于在现实世界的进一步验证。本综述总结了AI在青光眼领域的应用，讨论了AI在当前临床工作中的限制以及需要注意的事项。
[Key word]
[Abstract]
Currently, the early diagnosis of glaucoma and monitoring of disease progression is difficult and requires assessment of structural(fundus photo/ optical coherence tomography scan)and functional damage(visual fields)of the optic nerve head(ONH). It requires the clinical knowledge of glaucoma experts and is highly labor intensive. Artificial intelligence(AI)applications have been proposed to improve the understanding of glaucoma and help to reduce the time and manpower required for such clinical tasks. With the advent of deep learning(DL), many tools for ophthalmological image enhancement, segmentation and classification have also emerged. Especially in the last three years, a large number of algorithms suitable for analyzing the ONH structure and/or function, which have been proposed to help in glaucoma detection. AI tools have also been developed to predict the early progression of the disease. Bring the possibility of personalized precision treatment. However, these algorithms are yet to be tested in the real world. This review summarizes the diverse landscape of AI algorithms developed for glaucoma. We also discuss the current limitations and challenges that we need to overcome.
[中图分类号]
[基金项目]
国家自然科学基金项目(No.81960176)
