Automated quantification of anterior chamber cells using swept-source anterior segment optical coherence tomography.

IF 2.9 Q1 OPHTHALMOLOGY

Journal of Ophthalmic Inflammation and Infection Pub Date : 2025-01-09 DOI:10.1186/s12348-025-00456-y

Shani Pillar, Shin Kadomoto, Keren Chen, Saitiel Sandoval Gonzalez, Nina Cherian, Joseph K Privratsky, Nicolette Zargari, Nicholas J Jackson, Giulia Corradetti, Judy L Chen, SriniVas R Sadda, Gary N Holland, Edmund Tsui

{"title":"Automated quantification of anterior chamber cells using swept-source anterior segment optical coherence tomography.","authors":"Shani Pillar, Shin Kadomoto, Keren Chen, Saitiel Sandoval Gonzalez, Nina Cherian, Joseph K Privratsky, Nicolette Zargari, Nicholas J Jackson, Giulia Corradetti, Judy L Chen, SriniVas R Sadda, Gary N Holland, Edmund Tsui","doi":"10.1186/s12348-025-00456-y","DOIUrl":null,"url":null,"abstract":"Purpose: To validate automated counts of presumed anterior chamber (AC) cells in eyes with histories of uveitis involving the anterior segment using swept-source (SS) anterior segment optical coherence tomography (AS-OCT) against manual counts and compare automated counts against Standardized Uveitis Nomenclature (SUN) criteria.Methods: Eyes were imaged with the ANTERION SS AS-OCT device (Heidelberg Engineering). A fully automated custom algorithm quantified the number of hyper-reflective foci (HRF) in line-scan images. Automated and manual counts were compared using interclass correlation (ICC) and Pearson correlation coefficient. Automated counts were compared to SUN grades using a mixed-effects linear regression model.Results: 90 eyes (54 participants) were included; 67 eyes (41 participants) had histories of uveitis, while 23 eyes (13 healthy participants) served as controls. ICC comparing automated to manual counts was 0.99 and the Pearson correlation coefficient was 0.98. Eyes at each SUN grade with corresponding median HRF (interquartile range [IQR]) were: Grade 0, 42 eyes, 2 HRF (0,4); 0.5+, 10 eyes, 10 HRF (8,15); 1+, 9 eyes, 22 HRF (15,33); 2+, 3 eyes, 27 HRF; 3+, 2 eyes, 128 HRF; 4+, 1 eye, 474 HRF. For every 1-step increase in grade, automated count increased by 38 (p < 0.001) or 293% (Pearson correlation coefficient 0.80, p < 0.001). Automated counts differed significantly between clinically inactive eyes (2 HRF [0,4]) and controls (0 HRF [0,1], p = 0.02).Conclusions: Our algorithm accurately counts HRF when compared to manual counts, with strong correlation to SUN clinical grades. SS AS-OCT offers the advantage of imaging of the entire AC and may allow detection of subclinical inflammation in eyes that appear clinically inactive.","PeriodicalId":16600,"journal":{"name":"Journal of Ophthalmic Inflammation and Infection","volume":"15 1","pages":"3"},"PeriodicalIF":2.9000,"publicationDate":"2025-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11717729/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Ophthalmic Inflammation and Infection","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1186/s12348-025-00456-y","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"OPHTHALMOLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Purpose: To validate automated counts of presumed anterior chamber (AC) cells in eyes with histories of uveitis involving the anterior segment using swept-source (SS) anterior segment optical coherence tomography (AS-OCT) against manual counts and compare automated counts against Standardized Uveitis Nomenclature (SUN) criteria.

Methods: Eyes were imaged with the ANTERION SS AS-OCT device (Heidelberg Engineering). A fully automated custom algorithm quantified the number of hyper-reflective foci (HRF) in line-scan images. Automated and manual counts were compared using interclass correlation (ICC) and Pearson correlation coefficient. Automated counts were compared to SUN grades using a mixed-effects linear regression model.

Results: 90 eyes (54 participants) were included; 67 eyes (41 participants) had histories of uveitis, while 23 eyes (13 healthy participants) served as controls. ICC comparing automated to manual counts was 0.99 and the Pearson correlation coefficient was 0.98. Eyes at each SUN grade with corresponding median HRF (interquartile range [IQR]) were: Grade 0, 42 eyes, 2 HRF (0,4); 0.5+, 10 eyes, 10 HRF (8,15); 1+, 9 eyes, 22 HRF (15,33); 2+, 3 eyes, 27 HRF; 3+, 2 eyes, 128 HRF; 4+, 1 eye, 474 HRF. For every 1-step increase in grade, automated count increased by 38 (p < 0.001) or 293% (Pearson correlation coefficient 0.80, p < 0.001). Automated counts differed significantly between clinically inactive eyes (2 HRF [0,4]) and controls (0 HRF [0,1], p = 0.02).

Conclusions: Our algorithm accurately counts HRF when compared to manual counts, with strong correlation to SUN clinical grades. SS AS-OCT offers the advantage of imaging of the entire AC and may allow detection of subclinical inflammation in eyes that appear clinically inactive.

查看原文本刊更多论文

使用扫源前段光学相干断层扫描对前房细胞进行自动定量。

目的：验证使用扫描源（SS）前段光学相干断层扫描（AS-OCT）对有涉及前段葡萄膜炎病史的眼睛中假定的前房（AC）细胞的自动计数与人工计数的对比，并将自动计数与标准化葡萄膜炎命名法（SUN）标准进行比较。方法：采用ANTERION SS AS-OCT设备（Heidelberg Engineering）对眼睛进行成像。一个完全自动化的自定义算法量化了线扫描图像中超反射焦点（HRF）的数量。使用类间相关（ICC）和Pearson相关系数对自动计数和人工计数进行比较。使用混合效应线性回归模型将自动计数与SUN评分进行比较。结果：共纳入90只眼（54名受试者）；67只眼睛（41名参与者）有葡萄膜炎病史，而23只眼睛（13名健康参与者）作为对照。比较自动计数和人工计数的ICC为0.99，Pearson相关系数为0.98。每个SUN分级的眼具有相应的中位HRF（四分位间距[IQR]）： 0,42眼，2 HRF (0,4)；0.5+， 10眼，10 HRF (8,15)；1+， 9眼，22 HRF (15,33)；2+， 3眼，27 HRF；3+， 2眼，128 HRF；4+， 1眼，474 HRF。每增加1级，自动计数增加38 (p)。结论：与人工计数相比，我们的算法准确地计数HRF，与SUN临床分级有很强的相关性。SS AS-OCT提供了整个AC成像的优势，可以检测临床上表现不活跃的眼睛的亚临床炎症。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊