Scholarlink

Purpose <p>Adenoviral keratoconjunctivitis is the most common infectious disease in ophthalmology. Its clinical forms include epidemic keratoconjunctivitis (EKC) and pharyngoconjunctival fever (PCF). EKC can be complicated by multiple subepithelial infiltrates (MSI) of the cornea, which affect visual outcomes. As at present, artificial intelligence and machine learning are becoming increasingly important not only in the diagnostic imaging field but also in the clinical data science field, we investigated a predictive model for the development of corneal MSI in adenoviral keratoconjunctivitis.</p> Study design <p>Retrospective cohort study.</p> Methods <p>One hundred and forty cases of adenovirus keratoconjunctivitis diagnosed at Hinokuma Eye Clinic were enrolled. The dependent variable was corneal MSI, and the independent variables were adenovirus genotype, bilaterality, subconjunctival hemorrhage, eyelid swelling, conjunctival edema, conjunctival opacity, pseudomembrane, corneal epithelial findings, preauricular lymphadenopathy, eye discharge, lacrimation, eye pain, foreign body sensation, and itchy eye, which were analyzed by a categorical data analysis program (CATDAP).</p> Results <p>For single independent variables, corneal epithelial findings (Akaike information criterion, AIC=−6.46) and type 54 (AIC=−4.30) showed high predictive performance. In the combination of multiple independent variables, Type 56 or Type 37 with corneal epithelial damage also showed high predictive performance.</p> Conclusion <p>Predicting the occurrence of corneal MSI has until now been considered difficult. When a patient with EKC has marginal corneal epithelial or subepithelial opacities at initial presentation, it is presumed that the patient has a relatively high risk of developing MSI. Knowing the prevalent virus type in advance, would prove helpful in the diagnosis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluation of factors relating to the development of multiple subepithelial corneal infiltrates of adenoviral keratoconjunctivitis using categorical data analysis program

  • 期刊论文

  • 发表日期：

    2025年5月26日

查看全文
  • Nobuyoshi Kitaichi,
  • Ryosuke Dei,
  • Rikutaro Hinokuma,
  • Ippei Yoshikawa,
  • Miki Hiraoka,
  • Kazuo Nakajima

摘要

Purpose

Adenoviral keratoconjunctivitis is the most common infectious disease in ophthalmology. Its clinical forms include epidemic keratoconjunctivitis (EKC) and pharyngoconjunctival fever (PCF). EKC can be complicated by multiple subepithelial infiltrates (MSI) of the cornea, which affect visual outcomes. As at present, artificial intelligence and machine learning are becoming increasingly important not only in the diagnostic imaging field but also in the clinical data science field, we investigated a predictive model for the development of corneal MSI in adenoviral keratoconjunctivitis.

Study design

Retrospective cohort study.

Methods

One hundred and forty cases of adenovirus keratoconjunctivitis diagnosed at Hinokuma Eye Clinic were enrolled. The dependent variable was corneal MSI, and the independent variables were adenovirus genotype, bilaterality, subconjunctival hemorrhage, eyelid swelling, conjunctival edema, conjunctival opacity, pseudomembrane, corneal epithelial findings, preauricular lymphadenopathy, eye discharge, lacrimation, eye pain, foreign body sensation, and itchy eye, which were analyzed by a categorical data analysis program (CATDAP).

Results

For single independent variables, corneal epithelial findings (Akaike information criterion, AIC=−6.46) and type 54 (AIC=−4.30) showed high predictive performance. In the combination of multiple independent variables, Type 56 or Type 37 with corneal epithelial damage also showed high predictive performance.

Conclusion

Predicting the occurrence of corneal MSI has until now been considered difficult. When a patient with EKC has marginal corneal epithelial or subepithelial opacities at initial presentation, it is presumed that the patient has a relatively high risk of developing MSI. Knowing the prevalent virus type in advance, would prove helpful in the diagnosis.