{"title":"Role of perfusion-related metrics in visual field progression of primary open-angle and exfoliation glaucoma: a natural language processing approach.","authors":"Ayumu Akagi, Kaito Narimoto, Kanta Ueda, Noriko Koizumi, Naoki Okumura, Masaki Tanito","doi":"10.1007/s10384-025-01277-1","DOIUrl":null,"url":null,"abstract":"<p><strong>Purpose: </strong>To investigate the association between vascular factors affecting optic nerve perfusion and glaucoma progression in a clinical setting with systematic monitoring of blood pressure and intraocular pressure (IOP).</p><p><strong>Study design: </strong>Retrospective cohort study.</p><p><strong>Methods: </strong>This retrospective study analyzed 428 primary open-angle glaucoma (POAG; n=295) and exfoliation glaucoma (EXG; n=133) patients from Shimane University Hospital. We employed a validated Bidirectional Encoder Representations from Transformers (BERT)-based natural language processing system to extract clinical data from electronic medical records of 36,561 visits, including blood pressure measurements, IOP, and medication details across 216 clinical parameters. Mixed-effects regression models were employed to identify factors associated with visual field mean deviation (MD) and the rate of MD decline.</p><p><strong>Results: </strong>The cohort demonstrated consistent pressure management (median IOP: 13.0 mmHg; median systolic: 138.0 mmHg; median diastolic: 79.0 mmHg). Ocular perfusion pressure (OPP) showed no significant association with either visual field MD or the rate of MD decline. Significant predictors of worse visual field MD included age (coefficient: -0.28, 95% CI -0.33 to -0.23, p<0.001), history of filtration surgery (coefficient: -3.8, 95% CI -5.7 to -1.9, p<0.001), and number of glaucoma medications (coefficient: -0.19, 95% CI -0.35 to -0.030, p=0.020). Factors significantly associated with faster MD decline were age, initial MD severity, and medication burden.</p><p><strong>Conclusions: </strong>In a clinical setting with systematic monitoring, OPP provided limited prognostic value for glaucoma progression, while traditional clinical factors were stronger predictors. Our BERT-based approach demonstrates how artificial intelligence (AI)-powered data extraction enables comprehensive analyses of clinical records, offering new insights for routine ophthalmological care.</p>","PeriodicalId":14563,"journal":{"name":"Japanese Journal of Ophthalmology","volume":" ","pages":""},"PeriodicalIF":1.9000,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Japanese Journal of Ophthalmology","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1007/s10384-025-01277-1","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"OPHTHALMOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Purpose: To investigate the association between vascular factors affecting optic nerve perfusion and glaucoma progression in a clinical setting with systematic monitoring of blood pressure and intraocular pressure (IOP).
Study design: Retrospective cohort study.
Methods: This retrospective study analyzed 428 primary open-angle glaucoma (POAG; n=295) and exfoliation glaucoma (EXG; n=133) patients from Shimane University Hospital. We employed a validated Bidirectional Encoder Representations from Transformers (BERT)-based natural language processing system to extract clinical data from electronic medical records of 36,561 visits, including blood pressure measurements, IOP, and medication details across 216 clinical parameters. Mixed-effects regression models were employed to identify factors associated with visual field mean deviation (MD) and the rate of MD decline.
Results: The cohort demonstrated consistent pressure management (median IOP: 13.0 mmHg; median systolic: 138.0 mmHg; median diastolic: 79.0 mmHg). Ocular perfusion pressure (OPP) showed no significant association with either visual field MD or the rate of MD decline. Significant predictors of worse visual field MD included age (coefficient: -0.28, 95% CI -0.33 to -0.23, p<0.001), history of filtration surgery (coefficient: -3.8, 95% CI -5.7 to -1.9, p<0.001), and number of glaucoma medications (coefficient: -0.19, 95% CI -0.35 to -0.030, p=0.020). Factors significantly associated with faster MD decline were age, initial MD severity, and medication burden.
Conclusions: In a clinical setting with systematic monitoring, OPP provided limited prognostic value for glaucoma progression, while traditional clinical factors were stronger predictors. Our BERT-based approach demonstrates how artificial intelligence (AI)-powered data extraction enables comprehensive analyses of clinical records, offering new insights for routine ophthalmological care.
期刊介绍:
The Japanese Journal of Ophthalmology (JJO) was inaugurated in 1957 as a quarterly journal published in English by the Ophthalmology Department of the University of Tokyo, with the aim of disseminating the achievements of Japanese ophthalmologists worldwide. JJO remains the only Japanese ophthalmology journal published in English. In 1997, the Japanese Ophthalmological Society assumed the responsibility for publishing the Japanese Journal of Ophthalmology as its official English-language publication.
Currently the journal is published bimonthly and accepts papers from authors worldwide. JJO has become an international interdisciplinary forum for the publication of basic science and clinical research papers.