Pub Date : 2026-03-01Epub Date: 2025-12-18DOI: 10.1016/j.xops.2025.101041
Douglas R. da Costa MD , Dániel Unyi , Rafael Scherer MD, PhD , Rohit Muralidhar , Mario Luiz Ribeiro Monteiro MD, PhD , Felipe A. Medeiros MD, PhD
Purpose
To evaluate the performance of a deep learning (DL) model based on graph isomorphism networks (GINs) for detecting glaucomatous visual field defects on 24-2 standard automated perimetry (SAP) and to compare it against traditional diagnostic criteria, a dense neural network (NN) model, and a convolutional neural network (CNN) model.
Design
A cross-sectional retrospective study. Participants: 1874 reliable SAP tests (Humphrey Field Analyzer, Carl-Zeiss Meditec Inc.) from 1009 eyes of 676 patients.
Methods
Standard automated perimetry tests were classified as normal or abnormal due to glaucomatous damage by two glaucoma specialists, with adjudication by a third. A GIN architecture was developed to classify tests using full 54-point spatial SAP data modeled as graphs, with node features comprising sensitivity, total deviation, and pattern deviation values. The dataset was split at the patient level (60% training/validation, 40% testing). The GIN model’s diagnostic performance was compared to the Anderson criteria, the glaucoma hemifield test/pattern standard deviation (GHT/PSD) criteria, a fully connected dense NN, and a CNN model.
Main Outcome Measures
Area under the receiver operating characteristic curve (AUC), precision–recall curve, sensitivity at 95% specificity, F1-score, repeatability, and model explainability.
Results
Among the 1874 SAP tests, 70.0% were graded as abnormal. The GIN model achieved an AUC of 0.982, significantly outperforming the Anderson criteria (AUC: 0.906, P < 0.001), GHT/PSD (AUC: 0.936, P = 0.006), the NN model (AUC: 0.941, P = 0.007), and the CNN model (AUC: 0.941, P = 0.027). At 95% specificity, the GIN model reached the highest sensitivity of 94.1%, surpassing the NN model (88.3%), CNN model (92.0%), GHT/PSD (90.1%), and Anderson criteria (85.1%). The GIN model also achieved the highest average precision (0.952) among evaluated criteria. Explainability analysis using GraphNOSE demonstrated that the GIN model emphasized clinically relevant regions associated with glaucomatous loss, offering interpretability advantages over conventional DL approaches.
Conclusions
By modeling SAP as a graph and incorporating spatial relationships among test points, the GIN model provided superior diagnostic performance and interpretability relative to traditional criteria and standard NNs. This graph-based approach offers a promising tool for accurate and explainable detection of glaucomatous visual field defects in clinical practice.
Financial Disclosures
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
目的评价基于图同态网络（GINs）的深度学习（DL）模型在24-2标准自动视野测量（SAP）上检测青光眼视野缺陷的性能，并将其与传统诊断标准、密集神经网络（NN）模型和卷积神经网络（CNN）模型进行比较。设计横断面回顾性研究。参与者：1874个可靠的SAP测试（Humphrey Field Analyzer, Carl-Zeiss Meditec Inc.），来自676名患者的1009只眼睛。方法由两名青光眼专家将标准的自动视距检查分为青光眼损伤的正常或异常，并由第三名青光眼专家判定。开发了一个GIN架构，使用完整的54点空间SAP数据建模为图来对测试进行分类，节点特征包括灵敏度、总偏差和模式偏差值。数据集在患者层面被分割（60%用于训练/验证，40%用于测试）。将GIN模型的诊断性能与Anderson标准、青光眼半场测试/模式标准差（GHT/PSD）标准、全连接密集神经网络和CNN模型进行比较。主要结果测量：受试者工作特征曲线（AUC）下的面积、精密度-召回率曲线、95%特异性的灵敏度、f1评分、可重复性和模型可解释性。结果1874例SAP试验中，70.0%为异常。GIN模型的AUC为0.982，显著优于Anderson标准（AUC: 0.906, P < 0.001）、GHT/PSD （AUC: 0.936, P = 0.006）、NN模型（AUC: 0.941, P = 0.007）和CNN模型（AUC: 0.941, P = 0.027）。在95%的特异性下，GIN模型达到了94.1%的最高灵敏度，超过了NN模型（88.3%）、CNN模型（92.0%）、GHT/PSD（90.1%）和Anderson标准（85.1%）。GIN模型在评价标准中平均精度最高（0.952）。使用GraphNOSE进行的可解释性分析表明，GIN模型强调与青光眼丧失相关的临床相关区域，与传统DL方法相比具有可解释性优势。通过将SAP建模为图形，并结合测试点之间的空间关系，GIN模型相对于传统标准和标准神经网络具有更好的诊断性能和可解释性。这种基于图的方法为临床实践中青光眼视野缺陷的准确和可解释的检测提供了一种有前途的工具。财务披露专有或商业披露可在本文末尾的脚注和披露中找到。
{"title":"Convolutional Graph Isomorphism Network to Detect Glaucomatous Visual Field Defects","authors":"Douglas R. da Costa MD , Dániel Unyi , Rafael Scherer MD, PhD , Rohit Muralidhar , Mario Luiz Ribeiro Monteiro MD, PhD , Felipe A. Medeiros MD, PhD","doi":"10.1016/j.xops.2025.101041","DOIUrl":"10.1016/j.xops.2025.101041","url":null,"abstract":"<div><h3>Purpose</h3><div>To evaluate the performance of a deep learning (DL) model based on graph isomorphism networks (GINs) for detecting glaucomatous visual field defects on 24-2 standard automated perimetry (SAP) and to compare it against traditional diagnostic criteria, a dense neural network (NN) model, and a convolutional neural network (CNN) model.</div></div><div><h3>Design</h3><div>A cross-sectional retrospective study. Participants: 1874 reliable SAP tests (Humphrey Field Analyzer, Carl-Zeiss Meditec Inc.) from 1009 eyes of 676 patients.</div></div><div><h3>Methods</h3><div>Standard automated perimetry tests were classified as normal or abnormal due to glaucomatous damage by two glaucoma specialists, with adjudication by a third. A GIN architecture was developed to classify tests using full 54-point spatial SAP data modeled as graphs, with node features comprising sensitivity, total deviation, and pattern deviation values. The dataset was split at the patient level (60% training/validation, 40% testing). The GIN model’s diagnostic performance was compared to the Anderson criteria, the glaucoma hemifield test/pattern standard deviation (GHT/PSD) criteria, a fully connected dense NN, and a CNN model.</div></div><div><h3>Main Outcome Measures</h3><div>Area under the receiver operating characteristic curve (AUC), precision–recall curve, sensitivity at 95% specificity, F1-score, repeatability, and model explainability.</div></div><div><h3>Results</h3><div>Among the 1874 SAP tests, 70.0% were graded as abnormal. The GIN model achieved an AUC of 0.982, significantly outperforming the Anderson criteria (AUC: 0.906, <em>P</em> < 0.001), GHT/PSD (AUC: 0.936, <em>P</em> = 0.006), the NN model (AUC: 0.941, <em>P</em> = 0.007), and the CNN model (AUC: 0.941, <em>P</em> = 0.027). At 95% specificity, the GIN model reached the highest sensitivity of 94.1%, surpassing the NN model (88.3%), CNN model (92.0%), GHT/PSD (90.1%), and Anderson criteria (85.1%). The GIN model also achieved the highest average precision (0.952) among evaluated criteria. Explainability analysis using GraphNOSE demonstrated that the GIN model emphasized clinically relevant regions associated with glaucomatous loss, offering interpretability advantages over conventional DL approaches.</div></div><div><h3>Conclusions</h3><div>By modeling SAP as a graph and incorporating spatial relationships among test points, the GIN model provided superior diagnostic performance and interpretability relative to traditional criteria and standard NNs. This graph-based approach offers a promising tool for accurate and explainable detection of glaucomatous visual field defects in clinical practice.</div></div><div><h3>Financial Disclosures</h3><div>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101041"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079753","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-12-26DOI: 10.1016/j.xops.2025.101050
Sahana Srinivasan BEng , Xuguang Ai MSc , Thaddaeus Wai Soon Lo BEng , Aidan Gilson MD , Minjie Zou MD, MMed , Ke Zou PhD , Hyunjae Kim PhD , Mingjia Yang BEng , Krithi Pushpanathan MSc , Samantha Min Er Yew MSc , Wan Ting Loke MSc , Jocelyn Hui Lin Goh BEng , Yibing Chen BEng , Yiming Kong , Emily Yuelei Fu MSc , Michelle Ong BEng , Kristen Nwanyanwu MD, MHS , Amisha Dave MD , Kelvin Zhenghao Li MBBS, MMed , Chen-Hsin Sun MD. MMed , Yih-Chung Tham PhD
<div><h3>Purpose</h3><div>Current benchmarks evaluating large language models (LLMs) in ophthalmology are narrow and disproportionately prioritize accuracy. We introduce BEnchmarking LLMs for Ophthalmology (BELO), a standardized evaluation benchmark developed through multiple rounds of expert checking by 13 ophthalmologists. BEnchmarking LLMs for Ophthalmology assesses ophthalmology-related knowledge and reasoning quality.</div></div><div><h3>Subjects</h3><div>This study did not involve human participation.</div></div><div><h3>Design</h3><div>Cross-sectional study.</div></div><div><h3>Methods</h3><div>Using keyword matching and a fine-tuned PubMed Bidirectional Encoder Representations from Transformers model, we curated ophthalmology-specific multiple-choice questions (MCQs) from diverse medical data sets (Basic and Clinical Science Course [BCSC], Multi-Subject Multi-Choice Dataset for Medical domain [MedMCQA], Medical Question Answering [MedQA], Biomedical Semantic Indexing and Question Answering [BioASQ], and PubMed Question Answering [PubMedQA]). The data set underwent multiple rounds of expert checking. Duplicate and substandard questions were systematically removed. Ten ophthalmologists refined the explanations of each MCQ's correct answer. This was further adjudicated by 3 senior ophthalmologists. To illustrate BELO's utility, we evaluated 8 LLMs (OpenAI o1, o3-mini, GPT-5, GPT-4o, DeepSeek-R1, MedGemma-4B, Llama-3-8B, and Gemini 1.5 Pro).</div></div><div><h3>Main Outcome Measures</h3><div>The 8 LLMs were evaluated in terms using accuracy, macro-F1, and 5 text-generation metrics (Recall-Oriented Understudy for Gisting Evaluation, BERTScore, BARTScore, Metric for Evaluation of Translation with Explicit Ordering, and AlignScore). In a further evaluation involving human experts, 2 ophthalmologists qualitatively reviewed 50 randomly selected outputs for accuracy, comprehensiveness, and completeness.</div></div><div><h3>Results</h3><div>BEnchmarking LLMs for Ophthalmology consists of 900 high-quality, expert-reviewed questions aggregated from 5 sources: BCSC (260), BioASQ (10), MedMCQA (572), MedQA (40), and PubMedQA (18). To demonstrate BELO's utility, we conducted a series of benchmarking exercises. In the quantitative evaluation, GPT-5 achieved the highest accuracy (0.90, 95% confidence interval [CI]: 0.89–0.92) and macro-F1 score (0.91, 95% CI: 0.89–0.93). On the other hand, the models' performance on text-generation metrics varied and were generally suboptimal, with scores ranging from 20.4 to 72.0 (out of 100, excluding the BARTScore metric), indicating room for improvement in clinical reasoning. In expert evaluations, GPT-4o was rated highest for accuracy and readability, while Gemini 1.5 Pro scored highest for completeness. A public leaderboard has been established to promote transparent evaluation and reporting. Importantly, the BELO data set will remain a hold-out, evaluation-only benchmark to ensure fair and reproducible comparisons o
{"title":"BEnchmarking Large Language Models for Ophthalmology (BELO): An Expert-Curated Data Set and Evaluation Framework for Knowledge and Reasoning","authors":"Sahana Srinivasan BEng , Xuguang Ai MSc , Thaddaeus Wai Soon Lo BEng , Aidan Gilson MD , Minjie Zou MD, MMed , Ke Zou PhD , Hyunjae Kim PhD , Mingjia Yang BEng , Krithi Pushpanathan MSc , Samantha Min Er Yew MSc , Wan Ting Loke MSc , Jocelyn Hui Lin Goh BEng , Yibing Chen BEng , Yiming Kong , Emily Yuelei Fu MSc , Michelle Ong BEng , Kristen Nwanyanwu MD, MHS , Amisha Dave MD , Kelvin Zhenghao Li MBBS, MMed , Chen-Hsin Sun MD. MMed , Yih-Chung Tham PhD","doi":"10.1016/j.xops.2025.101050","DOIUrl":"10.1016/j.xops.2025.101050","url":null,"abstract":"<div><h3>Purpose</h3><div>Current benchmarks evaluating large language models (LLMs) in ophthalmology are narrow and disproportionately prioritize accuracy. We introduce BEnchmarking LLMs for Ophthalmology (BELO), a standardized evaluation benchmark developed through multiple rounds of expert checking by 13 ophthalmologists. BEnchmarking LLMs for Ophthalmology assesses ophthalmology-related knowledge and reasoning quality.</div></div><div><h3>Subjects</h3><div>This study did not involve human participation.</div></div><div><h3>Design</h3><div>Cross-sectional study.</div></div><div><h3>Methods</h3><div>Using keyword matching and a fine-tuned PubMed Bidirectional Encoder Representations from Transformers model, we curated ophthalmology-specific multiple-choice questions (MCQs) from diverse medical data sets (Basic and Clinical Science Course [BCSC], Multi-Subject Multi-Choice Dataset for Medical domain [MedMCQA], Medical Question Answering [MedQA], Biomedical Semantic Indexing and Question Answering [BioASQ], and PubMed Question Answering [PubMedQA]). The data set underwent multiple rounds of expert checking. Duplicate and substandard questions were systematically removed. Ten ophthalmologists refined the explanations of each MCQ's correct answer. This was further adjudicated by 3 senior ophthalmologists. To illustrate BELO's utility, we evaluated 8 LLMs (OpenAI o1, o3-mini, GPT-5, GPT-4o, DeepSeek-R1, MedGemma-4B, Llama-3-8B, and Gemini 1.5 Pro).</div></div><div><h3>Main Outcome Measures</h3><div>The 8 LLMs were evaluated in terms using accuracy, macro-F1, and 5 text-generation metrics (Recall-Oriented Understudy for Gisting Evaluation, BERTScore, BARTScore, Metric for Evaluation of Translation with Explicit Ordering, and AlignScore). In a further evaluation involving human experts, 2 ophthalmologists qualitatively reviewed 50 randomly selected outputs for accuracy, comprehensiveness, and completeness.</div></div><div><h3>Results</h3><div>BEnchmarking LLMs for Ophthalmology consists of 900 high-quality, expert-reviewed questions aggregated from 5 sources: BCSC (260), BioASQ (10), MedMCQA (572), MedQA (40), and PubMedQA (18). To demonstrate BELO's utility, we conducted a series of benchmarking exercises. In the quantitative evaluation, GPT-5 achieved the highest accuracy (0.90, 95% confidence interval [CI]: 0.89–0.92) and macro-F1 score (0.91, 95% CI: 0.89–0.93). On the other hand, the models' performance on text-generation metrics varied and were generally suboptimal, with scores ranging from 20.4 to 72.0 (out of 100, excluding the BARTScore metric), indicating room for improvement in clinical reasoning. In expert evaluations, GPT-4o was rated highest for accuracy and readability, while Gemini 1.5 Pro scored highest for completeness. A public leaderboard has been established to promote transparent evaluation and reporting. Importantly, the BELO data set will remain a hold-out, evaluation-only benchmark to ensure fair and reproducible comparisons o","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101050"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146189246","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-01-13DOI: 10.1016/j.xops.2026.101068
Rui Ning PhD , Yuechen Ren MS , Jinxuan Xiahou PhD , Shuoyu Xu MS , Kexin Li MS , Yiran Wang PhD , Xinning Yang MS , Zheng Li PhD , Xingtao Zhou PhD , Elsheikh Ahmed PhD , Xiaoying Wang PhD , Jinhai Huang PhD
Purpose
To assess Stress–Strain Index (SSI) map parameters and to establish a new corneal-biomechanics-based staging (CBBS) system and the diurnal variation of SSI.
Design
Hospital-based cross-sectional study.
Subjects
Seventy-eight keratoconus subjects.
Methods
A total of 78 keratoconus (KC) subjects were included in this study. All subjects had corneal tomography (Pentacam HR, Oculus) and biomechanical measurements (Corvis ST, Oculus) 5 times a day, which were at 8:30, 11:30, 14:30, 17:30, and 20:30. Stress–Strain Index value was obtained from Corvis ST, and SSI map parameters were derived from a customized SSI map generator (Matlab runtime 9.8). An analysis of variance was used to compare the SSI map parameters among varying stages of KC groups. The best performing parameters according to the Youden index were subsequently used to create the CBBS system. Cohen κ statistics and contingency tables were used to compare CBBS and topographic KC classification (TKC) systems.
Main Outcome Measures
Stress–Strain Index map parameters were used for analysis: mean inside cone SSI, mean outside cone SSI, minimum SSI, and maximum SSI.
Results
Significant diurnal variations were observed in mean inside cone SSI (F = 5.536, P < 0.001) and Min SSI (F = 6.031, P < 0.001) without any clinical significance. Mean inside cone SSI had the highest area under the curve value and sensitivity among the 3 KC stages and was used to establish the CBBS system. Cohen κ statistics (0.652, P < 0.001) and contingency tables (76.9% KC eyes were of the same stage) showed good agreement between CBBS and TKC systems.
Conclusions
A new corneal-biomechanics-based KC staging system was established on the basis of localized corneal biomechanics (mean inside cone SSI). There was no clinically significant diurnal variation in localized corneal biomechanics based on SSI map parameters.
Financial Disclosure(s)
The authors have no proprietary or commercial interest in any materials discussed in this article.
{"title":"Staging of Keratoconus Indices Based on Localized Corneal Biomechanics Stress–Strain Index Map","authors":"Rui Ning PhD , Yuechen Ren MS , Jinxuan Xiahou PhD , Shuoyu Xu MS , Kexin Li MS , Yiran Wang PhD , Xinning Yang MS , Zheng Li PhD , Xingtao Zhou PhD , Elsheikh Ahmed PhD , Xiaoying Wang PhD , Jinhai Huang PhD","doi":"10.1016/j.xops.2026.101068","DOIUrl":"10.1016/j.xops.2026.101068","url":null,"abstract":"<div><h3>Purpose</h3><div>To assess Stress–Strain Index (SSI) map parameters and to establish a new corneal-biomechanics-based staging (CBBS) system and the diurnal variation of SSI.</div></div><div><h3>Design</h3><div>Hospital-based cross-sectional study.</div></div><div><h3>Subjects</h3><div>Seventy-eight keratoconus subjects.</div></div><div><h3>Methods</h3><div>A total of 78 keratoconus (KC) subjects were included in this study. All subjects had corneal tomography (Pentacam HR, Oculus) and biomechanical measurements (Corvis ST, Oculus) 5 times a day, which were at 8:30, 11:30, 14:30, 17:30, and 20:30. Stress–Strain Index value was obtained from Corvis ST, and SSI map parameters were derived from a customized SSI map generator (Matlab runtime 9.8). An analysis of variance was used to compare the SSI map parameters among varying stages of KC groups. The best performing parameters according to the Youden index were subsequently used to create the CBBS system. Cohen κ statistics and contingency tables were used to compare CBBS and topographic KC classification (TKC) systems.</div></div><div><h3>Main Outcome Measures</h3><div>Stress–Strain Index map parameters were used for analysis: mean inside cone SSI, mean outside cone SSI, minimum SSI, and maximum SSI.</div></div><div><h3>Results</h3><div>Significant diurnal variations were observed in mean inside cone SSI (<em>F</em> = 5.536, <em>P</em> < 0.001) and Min SSI (<em>F</em> = 6.031, <em>P</em> < 0.001) without any clinical significance. Mean inside cone SSI had the highest area under the curve value and sensitivity among the 3 KC stages and was used to establish the CBBS system. Cohen κ statistics (0.652, <em>P</em> < 0.001) and contingency tables (76.9% KC eyes were of the same stage) showed good agreement between CBBS and TKC systems.</div></div><div><h3>Conclusions</h3><div>A new corneal-biomechanics-based KC staging system was established on the basis of localized corneal biomechanics (mean inside cone SSI). There was no clinically significant diurnal variation in localized corneal biomechanics based on SSI map parameters.</div></div><div><h3>Financial Disclosure(s)</h3><div>The authors have no proprietary or commercial interest in any materials discussed in this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101068"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147328495","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-01-06DOI: 10.1016/j.xops.2025.101062
Luning Yang PhD , Sen Lin MS , Yiwen Tao BS , Qi Pan MS , Tengda Cai MS , Yunyan Ye PhD , Jianhui Liu PhD , Yang Zhou PhD , Yongqing Shao MS , Quanyong Yi PhD , Zen Huat Lu PhD , Lie Chen PhD , Gareth McKay PhD , Richard Rankin PhD , Fan Li PhD , Weihua Meng PhD
Purpose
To characterize cell-type-specific transcriptional changes during human retinal aging and develop machine learning (ML) model for cellular age discrimination in a Chinese cohort.
Eighteen unfrozen retinas from 12 Chinese donors (9 young, 34-55 y; 9 old, 68-92 y).
Methods
Single-cell RNA sequencing (10x, v3.1) generated 223 612 cells, batch-corrected with single-cell variational inference; age-related signatures were defined by intersecting single-cell and pseudobulk differentially expressed genes (DEGs), then cell-type-specific panels were rank-ordered with L1-regularized logistic regression plus recursive feature elimination and interpreted through hallmark-pathway enrichment and transcription factor (TF) regulon mapping.
Main Outcome Measures
Age-related cellular composition shifts; cell-type-specific DEGs; ML classifier accuracy and feature rankings; TF regulon activity changes.
Results
Eleven major retinal cell populations were identified. Aging showed declining rod-to-cone ratios, reduced bipolar cell (BC) proportions among interneurons, and increased astrocyte abundance. Müller glial cells exhibited the most pronounced transcriptional changes, followed by BCs and rods. Machine learning classifiers achieved 80% to 96% accuracy across cell types (microglia 96%, horizontal cells [HCs] 93%, BCs 91%, cones 90%, rods 89%). Shared aging signatures included mitochondrial dysfunction and inflammatory activation. Cell-specific vulnerabilities emerged: mitochondria-centric stress in rods/BCs, proteostasis-retinoid metabolism in cones, and structural-RNA maintenance in HCs.
Conclusions
This study provides the first ML derived, cell-type-specific aging signatures for human retina in a Chinese cohort, revealing both conserved molecular hallmarks and distinctive cellular vulnerabilities that inform targeted therapeutic strategies for retinal aging.
Financial Disclosure(s)
The author has no/the authors have no proprietary or commercial interest in any materials discussed in this article.
目的：表征人类视网膜衰老过程中细胞类型特异性转录变化，并在中国队列中建立细胞年龄判别的机器学习（ML）模型。设计：横断面，实验室为基础的观察性研究。参与者：来自12名中国供体的18例未冷冻视网膜（9例年轻，34-55岁；9例老年，68-92岁）。方法：单细胞RNA测序（10x, v3.1）产生223 612个细胞，用单细胞变分推理进行批量校正；通过交叉单细胞和假体差异表达基因（DEGs）定义年龄相关特征，然后使用l1正则化逻辑回归加递归特征消除对细胞类型特异性面板进行排序，并通过标记通路富集和转录因子（TF）调控子定位进行解释。主要观察指标：与年龄相关的细胞组成变化；cell-type-specific度;机器学习分类器精度和特征排名；TF调节活性改变。结果：鉴定出11个主要的视网膜细胞群。衰老表现为杆锥比下降，中间神经元中双极细胞（BC）比例减少，星形胶质细胞丰度增加。mller胶质细胞表现出最明显的转录变化，其次是BCs和杆状细胞。机器学习分类器在细胞类型上的准确率达到80%到96%（小胶质细胞96%，水平细胞[hc] 93%， bc 91%，锥细胞90%，杆状细胞89%）。共同的衰老特征包括线粒体功能障碍和炎症激活。细胞特异性的脆弱性出现了：杆状细胞/BCs中的线粒体中心应激，锥体中的蛋白质固定-类视黄酮代谢，以及hc中的结构- rna维持。结论：本研究在中国人群中首次提供了ML衍生的细胞类型特异性衰老特征，揭示了保守的分子特征和独特的细胞脆弱性，为视网膜衰老的靶向治疗策略提供了信息。财务披露：作者在本文中讨论的任何材料中没有任何专有或商业利益。
{"title":"Interpretable Aging Signatures in Human Retinal Cell Types Revealed by Single-Cell RNA Sequencing and Sparse Logistic Regression","authors":"Luning Yang PhD , Sen Lin MS , Yiwen Tao BS , Qi Pan MS , Tengda Cai MS , Yunyan Ye PhD , Jianhui Liu PhD , Yang Zhou PhD , Yongqing Shao MS , Quanyong Yi PhD , Zen Huat Lu PhD , Lie Chen PhD , Gareth McKay PhD , Richard Rankin PhD , Fan Li PhD , Weihua Meng PhD","doi":"10.1016/j.xops.2025.101062","DOIUrl":"10.1016/j.xops.2025.101062","url":null,"abstract":"<div><h3>Purpose</h3><div>To characterize cell-type-specific transcriptional changes during human retinal aging and develop machine learning (ML) model for cellular age discrimination in a Chinese cohort.</div></div><div><h3>Design</h3><div>Cross-sectional, laboratory-based observational study.</div></div><div><h3>Participants</h3><div>Eighteen unfrozen retinas from 12 Chinese donors (9 young, 34-55 y; 9 old, 68-92 y).</div></div><div><h3>Methods</h3><div>Single-cell RNA sequencing (10x, v3.1) generated 223 612 cells, batch-corrected with single-cell variational inference; age-related signatures were defined by intersecting single-cell and pseudobulk differentially expressed genes (DEGs), then cell-type-specific panels were rank-ordered with L1-regularized logistic regression plus recursive feature elimination and interpreted through hallmark-pathway enrichment and transcription factor (TF) regulon mapping.</div></div><div><h3>Main Outcome Measures</h3><div>Age-related cellular composition shifts; cell-type-specific DEGs; ML classifier accuracy and feature rankings; TF regulon activity changes.</div></div><div><h3>Results</h3><div>Eleven major retinal cell populations were identified. Aging showed declining rod-to-cone ratios, reduced bipolar cell (BC) proportions among interneurons, and increased astrocyte abundance. Müller glial cells exhibited the most pronounced transcriptional changes, followed by BCs and rods. Machine learning classifiers achieved 80% to 96% accuracy across cell types (microglia 96%, horizontal cells [HCs] 93%, BCs 91%, cones 90%, rods 89%). Shared aging signatures included mitochondrial dysfunction and inflammatory activation. Cell-specific vulnerabilities emerged: mitochondria-centric stress in rods/BCs, proteostasis-retinoid metabolism in cones, and structural-RNA maintenance in HCs.</div></div><div><h3>Conclusions</h3><div>This study provides the first ML derived, cell-type-specific aging signatures for human retina in a Chinese cohort, revealing both conserved molecular hallmarks and distinctive cellular vulnerabilities that inform targeted therapeutic strategies for retinal aging.</div></div><div><h3>Financial Disclosure(s)</h3><div>The author has no/the authors have no proprietary or commercial interest in any materials discussed in this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101062"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146183798","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-01-17DOI: 10.1016/j.xops.2026.101077
Yeabsira Mesfin , Patrick Takla , Maanasa Indaram MD , Julius T. Oatts MD
Objective
Obtaining precise strabismus measurements is key in the diagnosis and management of strabismus in adults. Virtual reality (VR) has the potential to address the limitations of the standard alternate cover test (ACT) in measuring strabismus. The goal of this study was to evaluate the performance of a VR-simulated ACT in adults with strabismus.
Design
Prospective cohort study.
Participants
Eligible consecutive participants ≥18 years with manifest strabismus ≥5 prism diopters (PD) at either distance or near were prospectively enrolled.
Intervention
Each patient underwent a VR-simulated ACT with eye-tracking technology (Olleyes VisuALL) followed by an ACT by a masked pediatric ophthalmologist. Bland–Altman plots, intraclass correlation coefficient (ICC), and Spearman correlation were used to compare the limits of agreement (LOA) and association between the 2 examinations for the entire cohort as well as those with deviations ≥10 PD.
Main Outcome Measures
Level of agreement between VR and standard strabismus measurements obtained from the VR and ACT.
Results
Of the 61 participants enrolled, 56 completed the VR test at near and 50 at distance. At distance, VR measurements were 0.88 PD higher than the ACT (upper LOA: 19.3, lower LOA: –17.6). At near, VR measurements were 0.28 PD higher than the ACT (upper LOA: 26.1, lower LOA: –25.5). For patients with manifest strabismus ≥10 PD, VR measurements were 0.26 PD lower than the ACT at distance (upper LOA: 19.5, lower LOA: –20.1) and at 3.97 PD lower at near (upper LOA: 18.8, lower LOA: –2). Wider variability was observed in patients with larger deviations. Significant correlations between VR and ACT measurements were observed at distance greater than near (ICC = 0.78, P < 0.0001; ICC = 0.57, P < 0.001, respectively). Patients with deviations ≥10 PD exhibited similar agreement (ICC: 0.72, P < 0.0001; near: ICC = 0.64, P < 0.0001).
Conclusions
A VR-simulated ACT demonstrated good correlation to standard strabismus measurements at distance and moderate correlation at near. The VR device underestimated deviations with larger deviations and tended to overestimate in those with smaller deviations. Virtual reality with eye-tracking technology holds promise in the assessment of adults with strabismus.
Financial Disclosure(s)
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
目的获得准确的斜视测量值是成人斜视诊断和治疗的关键。虚拟现实（VR）有可能解决标准交替盖测试（ACT）在测量斜视方面的局限性。本研究的目的是评估vr模拟ACT在成人斜视患者中的表现。前瞻性队列研究。符合条件的连续受试者≥18岁，明显斜视≥5棱镜屈光度（PD）在远处或近处均可入选。干预每位患者都进行了一次使用眼动追踪技术（Olleyes VisuALL）的虚拟现实模拟ACT，随后由蒙面儿科眼科医生进行ACT。使用Bland-Altman图、类内相关系数（ICC）和Spearman相关来比较整个队列以及偏差≥10pd的两种检查之间的一致性限（LOA）和相关性。主要观察指标：VR与标准斜视测量值之间的一致程度。结果在61名参与者中，56人在近距离完成了虚拟现实测试，50人在远距离完成了虚拟现实测试。在远处，VR测量值比ACT高0.88 PD（上LOA: 19.3，下LOA: -17.6）。在接近时，VR测量值比ACT高0.28 PD（上LOA: 26.1，下LOA: -25.5）。对于明显斜视≥10 PD的患者，距离VR测量值比ACT低0.26 PD（上LOA: 19.5，下LOA: -20.1），近距离VR测量值比ACT低3.97 PD（上LOA: 18.8，下LOA: -2）。在偏差较大的患者中观察到更大的变异性。在距离大于近处观察到VR和ACT测量之间的显著相关性（ICC = 0.78, P < 0.0001; ICC = 0.57, P < 0.001）。偏差≥10 PD的患者表现出类似的一致性（ICC: 0.72, P < 0.0001；接近：ICC = 0.64, P < 0.0001）。结论vr模拟ACT与标准斜视测量值在远处有良好的相关性，在近处有中等相关性。对于较大的偏差，VR设备有低估偏差的倾向，对于较小的偏差，VR设备有高估偏差的倾向。带有眼球追踪技术的虚拟现实技术在评估成人斜视方面具有前景。财务披露专有或商业披露可在本文末尾的脚注和披露中找到。
{"title":"A Virtual Reality Simulated Alternate Cover Test to Evaluate Adults with Strabismus","authors":"Yeabsira Mesfin , Patrick Takla , Maanasa Indaram MD , Julius T. Oatts MD","doi":"10.1016/j.xops.2026.101077","DOIUrl":"10.1016/j.xops.2026.101077","url":null,"abstract":"<div><h3>Objective</h3><div>Obtaining precise strabismus measurements is key in the diagnosis and management of strabismus in adults. Virtual reality (VR) has the potential to address the limitations of the standard alternate cover test (ACT) in measuring strabismus. The goal of this study was to evaluate the performance of a VR-simulated ACT in adults with strabismus.</div></div><div><h3>Design</h3><div>Prospective cohort study.</div></div><div><h3>Participants</h3><div>Eligible consecutive participants ≥18 years with manifest strabismus ≥5 prism diopters (PD) at either distance or near were prospectively enrolled.</div></div><div><h3>Intervention</h3><div>Each patient underwent a VR-simulated ACT with eye-tracking technology (Olleyes VisuALL) followed by an ACT by a masked pediatric ophthalmologist. Bland–Altman plots, intraclass correlation coefficient (ICC), and Spearman correlation were used to compare the limits of agreement (LOA) and association between the 2 examinations for the entire cohort as well as those with deviations ≥10 PD.</div></div><div><h3>Main Outcome Measures</h3><div>Level of agreement between VR and standard strabismus measurements obtained from the VR and ACT.</div></div><div><h3>Results</h3><div>Of the 61 participants enrolled, 56 completed the VR test at near and 50 at distance. At distance, VR measurements were 0.88 PD higher than the ACT (upper LOA: 19.3, lower LOA: –17.6). At near, VR measurements were 0.28 PD higher than the ACT (upper LOA: 26.1, lower LOA: –25.5). For patients with manifest strabismus ≥10 PD, VR measurements were 0.26 PD lower than the ACT at distance (upper LOA: 19.5, lower LOA: –20.1) and at 3.97 PD lower at near (upper LOA: 18.8, lower LOA: –2). Wider variability was observed in patients with larger deviations. Significant correlations between VR and ACT measurements were observed at distance greater than near (ICC = 0.78, <em>P</em> < 0.0001; ICC = 0.57, <em>P</em> < 0.001, respectively). Patients with deviations ≥10 PD exhibited similar agreement (ICC: 0.72, <em>P</em> < 0.0001; near: ICC = 0.64, <em>P</em> < 0.0001).</div></div><div><h3>Conclusions</h3><div>A VR-simulated ACT demonstrated good correlation to standard strabismus measurements at distance and moderate correlation at near. The VR device underestimated deviations with larger deviations and tended to overestimate in those with smaller deviations. Virtual reality with eye-tracking technology holds promise in the assessment of adults with strabismus.</div></div><div><h3>Financial Disclosure(s)</h3><div>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101077"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146189732","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-01-14DOI: 10.1016/j.xops.2026.101073
Taiga Inooka MD, PhD , Ryo Tomita MD, PhD , Ayana Suzumura MD, PhD , Shota Fujikawa MD , Yuki Kimura MD , Taro Kominami MD, PhD , Tetsuhito Kojima MD, PhD , Shinji Ueno MD, PhD , Yasuki Ito MD, PhD , Koji M. Nishiguchi MD, PhD , Kenya Yuki MD, PhD
Purpose
To investigate the association between serum high-density lipoprotein cholesterol (HDL-C) levels and ganglion cell complex (GCC) thickness in a nonglaucomatous Japanese population.
Design
A retrospective cross-sectional observational study.
Participants
We included 588 nonglaucomatous Japanese adults who underwent comprehensive ophthalmic and systemic health screening.
Methods
Participants underwent OCT imaging, anthropometric measurements, including brachial–ankle pulse wave velocity, spirometry, and hematologic profiling. Multivariable linear regression models were used to assess the association between HDL-C levels and GCC thickness. Covariates were selected using a stepwise variable selection procedure, with the final model including age and axial length. A piecewise linear regression model further evaluated the association across different HDL-C ranges.
Main Outcome Measures
Average GCC thickness.
Results
Older age (P = 0.002), longer axial length (P < 0.001), and higher HDL-C levels (P < 0.001) were significantly associated with thinner GCC thickness. A nonlinear relationship was observed, with GCC thickness inversely associated with HDL-C levels outside the 60 to 67 mg/dL range (P = 0.005).
Conclusions
High-density lipoprotein cholesterol levels are significantly associated with GCC thickness in nonglaucomatous individuals, which suggests a potential role of lipid metabolism in early neuroretinal thinning. High-density lipoprotein cholesterol may serve as a biomarker for neurodegenerative changes, even before glaucomatous alterations become clinically apparent.
Financial Disclosure(s)
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
{"title":"Association of High-Density Lipoprotein Cholesterol with Macular Structure in Nonglaucomatous Individuals","authors":"Taiga Inooka MD, PhD , Ryo Tomita MD, PhD , Ayana Suzumura MD, PhD , Shota Fujikawa MD , Yuki Kimura MD , Taro Kominami MD, PhD , Tetsuhito Kojima MD, PhD , Shinji Ueno MD, PhD , Yasuki Ito MD, PhD , Koji M. Nishiguchi MD, PhD , Kenya Yuki MD, PhD","doi":"10.1016/j.xops.2026.101073","DOIUrl":"10.1016/j.xops.2026.101073","url":null,"abstract":"<div><h3>Purpose</h3><div>To investigate the association between serum high-density lipoprotein cholesterol (HDL-C) levels and ganglion cell complex (GCC) thickness in a nonglaucomatous Japanese population.</div></div><div><h3>Design</h3><div>A retrospective cross-sectional observational study.</div></div><div><h3>Participants</h3><div>We included 588 nonglaucomatous Japanese adults who underwent comprehensive ophthalmic and systemic health screening.</div></div><div><h3>Methods</h3><div>Participants underwent OCT imaging, anthropometric measurements, including brachial–ankle pulse wave velocity, spirometry, and hematologic profiling. Multivariable linear regression models were used to assess the association between HDL-C levels and GCC thickness. Covariates were selected using a stepwise variable selection procedure, with the final model including age and axial length. A piecewise linear regression model further evaluated the association across different HDL-C ranges.</div></div><div><h3>Main Outcome Measures</h3><div>Average GCC thickness.</div></div><div><h3>Results</h3><div>Older age (<em>P</em> = 0.002), longer axial length (<em>P</em> < 0.001), and higher HDL-C levels (<em>P</em> < 0.001) were significantly associated with thinner GCC thickness. A nonlinear relationship was observed, with GCC thickness inversely associated with HDL-C levels outside the 60 to 67 mg/dL range (<em>P</em> = 0.005).</div></div><div><h3>Conclusions</h3><div>High-density lipoprotein cholesterol levels are significantly associated with GCC thickness in nonglaucomatous individuals, which suggests a potential role of lipid metabolism in early neuroretinal thinning. High-density lipoprotein cholesterol may serve as a biomarker for neurodegenerative changes, even before glaucomatous alterations become clinically apparent.</div></div><div><h3>Financial Disclosure(s)</h3><div>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101073"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146189212","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-12-13DOI: 10.1016/j.xops.2025.101029
John D. Jackson MD , Mani K. Woodward MCR , David A. Sutter , Aaron S. Coyner PhD , Carol X. Wang , Susan R. Ostmo MS , Michael F. Chiang MD, MA , Yali Jia PhD , David Huang MD, PhD , Yifan Jian PhD , J. Peter Campbell MD, MPH , Benjamin K. Young MD, MS
Objective or Purpose
To develop a lightweight neural network for automated cross-sectional and en face segmentation of ultra-widefield (UWF) OCT images acquired for retinopathy of prematurity screening.
Design
Cross-sectional study.
Subjects
Twenty-five infants with a birth weight <1500 g or gestational age <31 weeks were scanned using a portable, handheld, swept-source UWF-OCT device.
Methods, Intervention, or Testing
For cross-sectional B-scans, 3040 B-scans from 5 OCT volumetric scans obtained from 5 patients were segmented by 2 graders for the choroid and retina using custom-built tools in the Napari image viewer. Using these segmentations, a u-net with an EfficientNet-B0 backbone was trained in combination with task-specific augmentations to perform automated segmentation of the retina and choroid data with varying levels of image processing applied. For en face scans, 40 en face images from 20 unique patients were manually segmented by a single grader for retinal vessels. Using these segmentations, a u-net with an EfficientNet-B0 backbone was trained. Validation for both B-scans and en face images was performed using fivefold cross-validation. The fivefold cross-validation metrics were then compared with the metrics obtained by comparing grader segmentations.
Main Outcome Measures
The Dice similarity coefficient (DSC) was used to assess B-scan and en face segmentations.
Results
The retinal and choroidal b-scan segmentations produced a DSC ± standard deviation of 0.925 ± 0.021 and 0.797 ± 0.062, respectively, averaged across the fivefolds. The en face vasculature segmentation produced a DSC ± standard deviation of 0.625 ± 0.0450.
Conclusions
Using u-net convolutional neural networks trained with task-specific augmentations, we developed en face and cross-sectional segmentations for UWF-OCT images, which will facilitate automated quantitative analysis with this novel modality.
Financial Disclosure(s)
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
{"title":"Automated Feature Segmentation of Ultra-Widefield OCT Images","authors":"John D. Jackson MD , Mani K. Woodward MCR , David A. Sutter , Aaron S. Coyner PhD , Carol X. Wang , Susan R. Ostmo MS , Michael F. Chiang MD, MA , Yali Jia PhD , David Huang MD, PhD , Yifan Jian PhD , J. Peter Campbell MD, MPH , Benjamin K. Young MD, MS","doi":"10.1016/j.xops.2025.101029","DOIUrl":"10.1016/j.xops.2025.101029","url":null,"abstract":"<div><h3>Objective or Purpose</h3><div>To develop a lightweight neural network for automated cross-sectional and en face segmentation of ultra-widefield (UWF) OCT images acquired for retinopathy of prematurity screening.</div></div><div><h3>Design</h3><div>Cross-sectional study.</div></div><div><h3>Subjects</h3><div>Twenty-five infants with a birth weight <1500 g or gestational age <31 weeks were scanned using a portable, handheld, swept-source UWF-OCT device.</div></div><div><h3>Methods, Intervention, or Testing</h3><div>For cross-sectional B-scans, 3040 B-scans from 5 OCT volumetric scans obtained from 5 patients were segmented by 2 graders for the choroid and retina using custom-built tools in the Napari image viewer. Using these segmentations, a u-net with an EfficientNet-B0 backbone was trained in combination with task-specific augmentations to perform automated segmentation of the retina and choroid data with varying levels of image processing applied. For en face scans, 40 en face images from 20 unique patients were manually segmented by a single grader for retinal vessels. Using these segmentations, a u-net with an EfficientNet-B0 backbone was trained. Validation for both B-scans and en face images was performed using fivefold cross-validation. The fivefold cross-validation metrics were then compared with the metrics obtained by comparing grader segmentations.</div></div><div><h3>Main Outcome Measures</h3><div>The Dice similarity coefficient (DSC) was used to assess B-scan and en face segmentations.</div></div><div><h3>Results</h3><div>The retinal and choroidal b-scan segmentations produced a DSC ± standard deviation of 0.925 ± 0.021 and 0.797 ± 0.062, respectively, averaged across the fivefolds. The en face vasculature segmentation produced a DSC ± standard deviation of 0.625 ± 0.0450.</div></div><div><h3>Conclusions</h3><div>Using u-net convolutional neural networks trained with task-specific augmentations, we developed en face and cross-sectional segmentations for UWF-OCT images, which will facilitate automated quantitative analysis with this novel modality.</div></div><div><h3>Financial Disclosure(s)</h3><div>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101029"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039784","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-12-29DOI: 10.1016/j.xops.2025.101040
Nianjia Wang , Xindi Liu , Liang Yao
{"title":"Re: Hiraoka et al: Quantitative Mapping of Posterior Eye Curvature in Children Using Distortion-Corrected OCT: Insights into Temporal Region Morphology","authors":"Nianjia Wang , Xindi Liu , Liang Yao","doi":"10.1016/j.xops.2025.101040","DOIUrl":"10.1016/j.xops.2025.101040","url":null,"abstract":"","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101040"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145969431","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
<div><h3>Purpose</h3><div>To identify genetic variants in posterior staphylomas in eyes with pathologic myopia using whole exome sequencing and to determine possible molecular mechanisms contributing to the pathogenesis.</div></div><div><h3>Design</h3><div>An observational, case-control study.</div></div><div><h3>Participants</h3><div>Two hundred sixty-four unrelated Japanese patients with myopia (≤ –0.50 diopters) and posterior staphyloma, which was diagnosed by ultra-widefield OCT, 3-dimensional magnetic resonance imaging, and Optos imaging.</div></div><div><h3>Methods</h3><div>Whole exome sequencing was performed on genomic DNA from peripheral blood. After variant filtering, the allelic frequencies were compared with control data obtained from East Asian subsets of the 1000 Genomes Project Phase III, the Exome Aggregation Consortium, and the Japanese Multi-Omics Reference Panel using Fisher exact test. A gene panel was constructed based on 13 staphyloma-associated disorders. Variants showing significant frequency differences (<em>P</em> ≤ 0.05) and an overlap of the gene panel were analyzed using gene set enrichment analysis with the DAVID Knowledgebase (v2023q4). Protein–protein interaction analysis was performed to assess functional associations.</div></div><div><h3>Main Outcome Measures</h3><div>The statistically associated variants and genes, gene set enrichment analysis results, protein–protein interaction networks, and involvement of basement membrane structures, including the inner limiting membrane (ILM) and Bruch membrane, were studied.</div></div><div><h3>Results</h3><div>Whole exome sequencing identified 16 656 missense variants in 8628 genes. Comparative allele frequency analyses with public databases revealed 3925 variants that had significantly higher allelic frequencies in the subjects. Of these, 81 genes overlapped with a curated staphyloma-related gene panel and were subjected to gene set enrichment analysis. The findings showed enrichment in basement membrane, extracellular matrix, and collagen-related pathways. The <em>COL4A5, COL18A1, COL2A1</em>, and <em>COL9A3</em> genes are concurrently enriched across these pathways. A missense variant in <em>COL4A5</em> was identified in 27 patients, and 96.3% of whom were females. Protein–protein interaction analysis demonstrated functional connections among these 4 genes.</div></div><div><h3>Conclusions</h3><div>Variants in the <em>COL4A5, COL18A1, COL2A1, and COL9A3</em> genes probably contribute to the pathogenesis of a posterior staphyloma through the disruption of collagen synthesis and basement membrane integrity. This was especially effective for the ILM and Bruch membrane. The <em>COL4A5</em> variant may cause an ocular-predominant phenotype in heterozygous female carriers, independent of the classical features of Alport syndrome.</div></div><div><h3>Financial Disclosure(s)</h3><div>The authors have no proprietary or commercial interest in any materials discussed in this artic
{"title":"Determining Genetic Cause of Posterior Staphylomas in Eyes with Pathologic Myopia by Whole Exome Sequencing","authors":"Ziye Wang MD , Changyu Chen MD, PhD , Yijin Wu MD , Yuki Nagata PhD , Toshihiro Tanaka MD, PhD , Shiqi Xie MD, PhD , Hongshuang Lu MD, PhD , Yining Wang MD , Jianping Xiong MD, PhD , Liwen Zhang MD , Koju Kamoi MD, PhD , Kyoko Ohno-Matsui MD, PhD","doi":"10.1016/j.xops.2025.101058","DOIUrl":"10.1016/j.xops.2025.101058","url":null,"abstract":"<div><h3>Purpose</h3><div>To identify genetic variants in posterior staphylomas in eyes with pathologic myopia using whole exome sequencing and to determine possible molecular mechanisms contributing to the pathogenesis.</div></div><div><h3>Design</h3><div>An observational, case-control study.</div></div><div><h3>Participants</h3><div>Two hundred sixty-four unrelated Japanese patients with myopia (≤ –0.50 diopters) and posterior staphyloma, which was diagnosed by ultra-widefield OCT, 3-dimensional magnetic resonance imaging, and Optos imaging.</div></div><div><h3>Methods</h3><div>Whole exome sequencing was performed on genomic DNA from peripheral blood. After variant filtering, the allelic frequencies were compared with control data obtained from East Asian subsets of the 1000 Genomes Project Phase III, the Exome Aggregation Consortium, and the Japanese Multi-Omics Reference Panel using Fisher exact test. A gene panel was constructed based on 13 staphyloma-associated disorders. Variants showing significant frequency differences (<em>P</em> ≤ 0.05) and an overlap of the gene panel were analyzed using gene set enrichment analysis with the DAVID Knowledgebase (v2023q4). Protein–protein interaction analysis was performed to assess functional associations.</div></div><div><h3>Main Outcome Measures</h3><div>The statistically associated variants and genes, gene set enrichment analysis results, protein–protein interaction networks, and involvement of basement membrane structures, including the inner limiting membrane (ILM) and Bruch membrane, were studied.</div></div><div><h3>Results</h3><div>Whole exome sequencing identified 16 656 missense variants in 8628 genes. Comparative allele frequency analyses with public databases revealed 3925 variants that had significantly higher allelic frequencies in the subjects. Of these, 81 genes overlapped with a curated staphyloma-related gene panel and were subjected to gene set enrichment analysis. The findings showed enrichment in basement membrane, extracellular matrix, and collagen-related pathways. The <em>COL4A5, COL18A1, COL2A1</em>, and <em>COL9A3</em> genes are concurrently enriched across these pathways. A missense variant in <em>COL4A5</em> was identified in 27 patients, and 96.3% of whom were females. Protein–protein interaction analysis demonstrated functional connections among these 4 genes.</div></div><div><h3>Conclusions</h3><div>Variants in the <em>COL4A5, COL18A1, COL2A1, and COL9A3</em> genes probably contribute to the pathogenesis of a posterior staphyloma through the disruption of collagen synthesis and basement membrane integrity. This was especially effective for the ILM and Bruch membrane. The <em>COL4A5</em> variant may cause an ocular-predominant phenotype in heterozygous female carriers, independent of the classical features of Alport syndrome.</div></div><div><h3>Financial Disclosure(s)</h3><div>The authors have no proprietary or commercial interest in any materials discussed in this artic","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101058"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-12-12DOI: 10.1016/j.xops.2025.101035
Joseph P.M. Blair PhD , Robyn H. Guymer MBBS, PhD , Alicja Krzemińska-Ściga MSc , Sandro De Zanet PhD , Carlos Ciller PhD , Stefanos Apostolopoulos PhD , Zhichao Wu BAppSc(Optom), PhD
Purpose
To examine the association between the loss of the OCT outer retinal bands and deep visual sensitivity losses quantified by defect-mapping microperimetry (DMP).
Design
Cross-sectional study.
Participants
Fifty individuals with geographic atrophy (GA) secondary to age-related macular degeneration.
Methods
All participants underwent DMP testing—a strategy optimized to quantify the spatial extent of deep visual sensitivity losses through single presentations of 10 decibel stimuli—with 208 locations sampled within the central 8° radius region. Participants also underwent OCT and fundus autofluorescence (FAF) imaging. OCT scans were automatically segmented to detect regions of retinal pigment epithelium (RPE), ellipsoid zone (EZ), and external limiting membrane (ELM) loss, and FAF images were manually annotated for GA. The extent of these parameters in the central 8° radius region where DMP testing was performed, and at each individual test location, was derived through image coregistration for evaluating structure-function associations.
Main Outcome Measures
Global structure-function correlation between the proportion of locations missed on DMP testing and the extent of loss of the structural parameters based on Spearman rank correlation coefficient (ρ), and spatial agreement between the presence of structural changes and missed stimuli on DMP testing at individual test locations based on the Dice similarity coefficient (DSC).
Results
There were strong global structure-function correlations based on loss of the OCT outer retinal bands (ρ = 0.85–0.86), similar to what was seen with FAF-defined GA (ρ = 0.89; P ≥ 0.326). However, the spatial agreement between OCT-defined EZ and ELM loss with missed stimuli on DMP testing (DSC = 0.64 for both) was higher than that seen with RPE loss (DSC = 0.60) and FAF-defined GA (DSC = 0.62; P = 0.008 for both), but the structure-function spatial agreement was similar between RPE loss and FAF-defined GA (P = 0.152).
Conclusions
Spatial agreement with pointwise deep visual sensitivity losses was comparable based on OCT-defined RPE loss and FAF-defined GA, but higher based on EZ and ELM loss. These findings confirm the expected functional relevance of these automatically derived OCT-defined parameters and support their utility as tools for monitoring GA progression.
Financial Disclosure(s)
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
{"title":"Geographic Atrophy Structure-Function Relationships Based on Loss of OCT Outer Retinal Bands and Fundus Autofluorescence","authors":"Joseph P.M. Blair PhD , Robyn H. Guymer MBBS, PhD , Alicja Krzemińska-Ściga MSc , Sandro De Zanet PhD , Carlos Ciller PhD , Stefanos Apostolopoulos PhD , Zhichao Wu BAppSc(Optom), PhD","doi":"10.1016/j.xops.2025.101035","DOIUrl":"10.1016/j.xops.2025.101035","url":null,"abstract":"<div><h3>Purpose</h3><div>To examine the association between the loss of the OCT outer retinal bands and deep visual sensitivity losses quantified by defect-mapping microperimetry (DMP).</div></div><div><h3>Design</h3><div>Cross-sectional study.</div></div><div><h3>Participants</h3><div>Fifty individuals with geographic atrophy (GA) secondary to age-related macular degeneration.</div></div><div><h3>Methods</h3><div>All participants underwent DMP testing—a strategy optimized to quantify the spatial extent of deep visual sensitivity losses through single presentations of 10 decibel stimuli—with 208 locations sampled within the central 8° radius region. Participants also underwent OCT and fundus autofluorescence (FAF) imaging. OCT scans were automatically segmented to detect regions of retinal pigment epithelium (RPE), ellipsoid zone (EZ), and external limiting membrane (ELM) loss, and FAF images were manually annotated for GA. The extent of these parameters in the central 8° radius region where DMP testing was performed, and at each individual test location, was derived through image coregistration for evaluating structure-function associations.</div></div><div><h3>Main Outcome Measures</h3><div>Global structure-function correlation between the proportion of locations missed on DMP testing and the extent of loss of the structural parameters based on Spearman rank correlation coefficient (ρ), and spatial agreement between the presence of structural changes and missed stimuli on DMP testing at individual test locations based on the Dice similarity coefficient (DSC).</div></div><div><h3>Results</h3><div>There were strong global structure-function correlations based on loss of the OCT outer retinal bands (ρ = 0.85–0.86), similar to what was seen with FAF-defined GA (ρ = 0.89; <em>P</em> ≥ 0.326). However, the spatial agreement between OCT-defined EZ and ELM loss with missed stimuli on DMP testing (DSC = 0.64 for both) was higher than that seen with RPE loss (DSC = 0.60) and FAF-defined GA (DSC = 0.62; <em>P</em> = 0.008 for both), but the structure-function spatial agreement was similar between RPE loss and FAF-defined GA (<em>P</em> = 0.152).</div></div><div><h3>Conclusions</h3><div>Spatial agreement with pointwise deep visual sensitivity losses was comparable based on OCT-defined RPE loss and FAF-defined GA, but higher based on EZ and ELM loss. These findings confirm the expected functional relevance of these automatically derived OCT-defined parameters and support their utility as tools for monitoring GA progression.</div></div><div><h3>Financial Disclosure(s)</h3><div>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</div></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"6 3","pages":"Article 101035"},"PeriodicalIF":4.6,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146079754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号