Revealing Academic Evolution and Frontier Pattern in the Field of Uveitis Using Bibliometric Analysis, Natural Language Processing, and Machine Learning

Numerous uveitis articles were published in this century, underneath which hides valuable intelligence. We aimed to characterize the evolution and patterns in this field. We divided the 15,994 uveitis papers into four consecutive time periods for bibliometric analysis, and applied latent Dirichlet allocation topic modeling and machine learning techniques to the latest period.  The yearly publication pattern fitted the curve: 1.21335x² − 4,848.95282x + 4,844,935.58876 (<i>R²</i> = 0.98311). The USA, the most productive country/region, focused on topics like ankylosing spondylitis and biologic therapy, whereas China (mainland) focused on topics like OCT and Behcet disease. The logistic regression showed the highest accuracy (71.6%) in the test set. In this century, a growing number of countries/regions/authors/journals are involved in the uveitis study, promoting the scientific output and thematic evolution. Our pioneering study uncovers the evolving academic trends and frontier patterns in this field using bibliometric analysis and AI algorithms.

本世纪已发表大量葡萄膜炎（uveitis）相关学术论文，其中蕴含着极具价值的研究情报。本研究旨在刻画该领域的发展演化历程与研究格局。我们将15994篇葡萄膜炎论文划分为四个连续的时间阶段开展文献计量分析，并针对最新阶段采用潜在狄利克雷分配（latent Dirichlet allocation）主题模型与机器学习技术进行分析。年度发文量拟合曲线为：1.21335x² − 4848.95282x + 4844935.58876（R²=0.98311）。美国是该领域发文量最高的国家/地区，其研究聚焦于强直性脊柱炎（ankylosing spondylitis）与生物治疗等主题；而中国大陆则围绕光学相干断层扫描（optical coherence tomography, OCT）、白塞病（Behcet disease）等方向开展研究。逻辑回归（logistic regression）在测试集上实现了最高的分类准确率，达71.6%。本世纪以来，参与葡萄膜炎研究的国家/地区、作者与期刊数量持续增长，推动了学术产出与主题演化。本研究开创性地通过文献计量分析与人工智能算法，揭示了该领域不断演进的学术趋势与前沿研究格局。