Advances in ophthalmology practice and research最新文献

{"title":"A systematic review of vision and vision-language foundation models in ophthalmology","authors":"Kai Jin ,&nbsp;Tao Yu ,&nbsp;Gui-shuang Ying ,&nbsp;Zongyuan Ge ,&nbsp;Kelvin Zhenghao Li ,&nbsp;Yukun Zhou ,&nbsp;Danli Shi ,&nbsp;Meng Wang ,&nbsp;Polat Goktas ,&nbsp;Andrzej Grzybowski","doi":"10.1016/j.aopr.2025.10.004","DOIUrl":"10.1016/j.aopr.2025.10.004","url":null,"abstract":"<div><h3>Background</h3><div>Vision and vision-language foundation models, a subset of advanced artificial intelligence (AI) frameworks, have shown transformative potential in various medical fields. In ophthalmology, these models, particularly large language models and vision-based models, have demonstrated great potential to improve diagnostic accuracy, enhance treatment planning, and streamline clinical workflows. However, their deployment in ophthalmology has faced several challenges, particularly regarding generalizability and integration into clinical practice. This systematic review aims to summarize the current evidence on the use of vision and vision-language foundation models in ophthalmology, identifying key applications, outcomes, and challenges.</div></div><div><h3>Main text</h3><div>A comprehensive search on PubMed, Web of Science, Scopus, and Google Scholar was conducted to identify studies published between January 2020 and July 2025. Studies were included if they developed or applied foundation models, such as vision-based models and large language models, to clinically relevant ophthalmic applications. A total of 10 studies met the inclusion criteria, covering areas such as retinal diseases, glaucoma, and ocular surface tumor. The primary outcome measures are model performance metrics, integration into clinical workflows, and the clinical utility of the models. Additionally, the review explored the limitations of foundation models, such as the reliance on large datasets, computational resources, and interpretability challenges.</div><div>The majority of studies demonstrated that foundation models could achieve high diagnostic accuracy, with several reports indicating excellent performance comparable to or exceeding those of experienced clinicians. Foundation models achieved high accuracy rates up to 95% for diagnosing retinal diseases, and similar performances for detecting glaucoma progression. Despite promising results, concerns about algorithmic bias, overfitting, and the need for diverse training data were common. High computational demands, EHR compatibility, and the need for clinician validation also posed challenges. Additionally, model interpretability issues hindered clinician trust and adoption.</div></div><div><h3>Conclusions</h3><div>Vision and vision-language foundation models in ophthalmology show significant potential for advancing diagnostic accuracy and treatment strategies, particularly in retinal diseases, glaucoma, and ocular oncology. However, challenges such as data quality, transparency, and ethical considerations must be addressed. Future research should focus on refining model performance, improving interpretability and generalizability, and exploring strategies for integrating these models into routine clinical practice to maximize their impact in clinical ophthalmology.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 8-19"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145799559","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}

{"title":"Mihe (MIOK) KPro with auricular cartilage reinforcement in end-stage corneal blindness: An open-label, nonrandomized trial","authors":"Qun Wang,&nbsp;Yuqi Wang,&nbsp;Mei Ge,&nbsp;Anqi Liu,&nbsp;Weiwei Xu,&nbsp;Liqiang Wang,&nbsp;Yifei Huang","doi":"10.1016/j.aopr.2025.11.001","DOIUrl":"10.1016/j.aopr.2025.11.001","url":null,"abstract":"<div><h3>Objective</h3><div>To evaluate efficacy and safety of the Mihe keratoprosthesis (MIOK) with autologous auricular cartilage reinforcement for bilateral end-stage corneal blindness.</div></div><div><h3>Methods</h3><div>This was a prospective, single-center, single-arm, non-randomized clinical trial. Patients with bilateral end-stage corneal blindness for which conventional corneal surgeries or standard Boston type I implantation were contraindicated, were enrolled. All participants underwent Mihe Kpro implantation through a two-stage surgical procedure. Autologous auricular cartilage was utilized to reinforce the tissue on the cornea's surface during each stage. Main outcome measures were keratoprosthesis retention rate, best corrected visual acuity (BCVA) and complications.</div></div><div><h3>Results</h3><div>Thirty-six patients were enrolled, and 35 of these patients completed the Mihe (MIOK) Kpro implantation. The most common indication for Mihe Kpro implantation was alkali burns (15 eyes, 42.9%), followed by end-stage dry eyes (11 eyes, 31.4%), melting metal burns (5 eyes, 14.3%), explosion injury (2 eyes, 5.71%), sulfuric acid burn (1 eyes, 2.86%), and multiple failed penetrating keratoplasty (1 eye, 2.86%). Thirty-four patients retained their initial keratoprosthesis until the final visit, with an average follow-up period of 68.84 ± 35.28 months (Median 85.55 months). Postoperative BCVA improved to 20/200 or better in 92% (23/25) of eyes at the final follow-up. Postoperative complications included fibrovascular tissue overgrowth on the optical surface (5 eyes, 14.3%), macular edema (5 eyes, 14.3%), residual corticle lens (4 eyes, 11.4%), newly acquired glaucoma (3 eyes, 8.57%), infectious endophthalmitis (3 eyes, 8.57%), transient intraocular pressure (IOP) elevation (2 eyes, 5.71%), retroprosthetic membrane (1 eye, 2.86%), and sterile vitritis (1 eye, 2.86%).</div></div><div><h3>Conclusions</h3><div>Mihe KPro reinforced with autologous auricular cartilage provides an effective and safe approach for visual rehabilitation in patients with bilateral end-stage corneal blindness.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 80-86"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146168057","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}

{"title":"RUNX1 promotes pathological retinal angiogenesis through von Willebrand factor","authors":"Ye Liu ,&nbsp;Chenfeng He ,&nbsp;Xiaogang Luo ,&nbsp;Ailing Sui ,&nbsp;Jingna He ,&nbsp;Yating Zhang ,&nbsp;Di Wu","doi":"10.1016/j.aopr.2025.11.006","DOIUrl":"10.1016/j.aopr.2025.11.006","url":null,"abstract":"<div><h3>Purpose</h3><div>Runt-related transcription factor 1 (RUNX1) is a key transcriptional regulator for angiogenesis, and its pharmacological inhibition attenuates retinal pathological angiogenesis. However, the functional mechanisms by which RUNX1 promotes angiogenesis, especially at the protein level, remain poorly understood.</div></div><div><h3>Methods</h3><div>Using an oxygen-induced retinopathy (OIR) model and a RUNX1 inhibition (Ro5-3335) injection model, we tested the effects of RUNX1 inhibition on neovascular formation. Retinas from mice were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) in Data-Independent Acquisition (DIA) mode. In vitro experiments were conducted to assess the effects of von Willebrand factor (vWF) on the migration and sprouting of RUNX1 overexpression endothelial cells (RUNX1-OE).</div></div><div><h3>Results</h3><div>In total, 465 differentially expressed proteins (DEPs) were identified, comprising 295 up-regulated proteins and 170 down-regulated proteins. The bioinformatic analysis revealed significant enrichment in pathways critical to retinal angiogenesis, including extracellular matrix (ECM)-receptor interaction, focal adhesion, and actin cytoskeleton regulation. Following RUNX1 inhibition, we identified 57 overlapping DEPs whose dysregulation was reversed. These rescued proteins showed strong enrichment in ECM-receptor interaction signaling. Notably, vWF, an endothelial-related marker, was identified as a downstream target protein of RUNX1 and was found to be involved in retinal pathological angiogenesis.</div></div><div><h3>Conclusions</h3><div>Our present findings establish RUNX1 as a promising therapeutic target for retinal pathological angiogenesis, with its effect mediated via the vWF axis.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 95-102"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146230009","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}

{"title":"The influence of a modified haptic on preventing toric IOL rotation in high rotation risk eyes","authors":"Shuyu Liu ,&nbsp;Chao Chen ,&nbsp;Jitong Zhou ,&nbsp;Kaiwen Cheng ,&nbsp;Yu Du ,&nbsp;Jiaqi Meng ,&nbsp;Yi Lu ,&nbsp;Wenwen He ,&nbsp;Xiangjia Zhu","doi":"10.1016/j.aopr.2025.11.003","DOIUrl":"10.1016/j.aopr.2025.11.003","url":null,"abstract":"<div><h3>Purpose</h3><div>To evaluate the rotational stability and astigmatic correction of a C-loop frosted haptic toric intraocular lens (IOL) compared with a plate-haptic toric IOL with/without capsular tension ring (CTR) implantation in cataractous eyes with high rotation risk due to large white-to-white (WTW) distance.</div></div><div><h3>Methods</h3><div>This retrospective cohort study included 90 cataractous eyes with WTW &gt; 11.8 mm, which received implantation of either a C-loop frosted haptic toric IOL (Group ZCU), a plate-haptic toric IOL (Group 709), or a plate-haptic toric IOL with CTR (Group 709 + CTR). IOL rotation and residual astigmatism were assessed one month postoperatively.</div></div><div><h3>Results</h3><div>Group ZCU demonstrated significantly better rotational stability, with a mean rotation of 1.87° ± 1.03°, compared to 4.63° ± 2.12° in Group 709 and 4.13° ± 1.86° in Group 709 + CTR (both <em>P</em> &lt; 0.05). Residual astigmatism was also significantly lower in Group ZCU (0.36 ± 0.26 D), compared to Group 709 (0.54 ± 0.43 D) and Group 709 + CTR (0.65 ± 0.27 D; both <em>P</em> &lt; 0.05). A positive correlation between WTW and IOL rotation was found in Group 709 (<em>r</em> = 0.419, <em>P</em> = 0.021) and Group 709 + CTR (<em>r</em> = 0.403, <em>P</em> = 0.027), but not in Group ZCU (<em>r</em> = −0.180, <em>P</em> = 0.341). Lens thickness and WTW were identified as independent predictors of rotation in Group 709, whereas only WTW was significant in Group 709 + CTR.</div></div><div><h3>Conclusions</h3><div>In eyes with high rotation risk due to large WTW (&gt; 11.8 mm), the ZCU IOL exhibits superior rotational stability and astigmatic correction compared to the 709 IOL, regardless of whether CTR is used.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 50-55"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146078130","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}

{"title":"Predicting intravitreal treatment response using ultrawide-field angiographic biomarkers in diabetic retinopathy","authors":"Karthik Reddy ,&nbsp;Callie Deng ,&nbsp;Boonkit Purt ,&nbsp;Yue Liang ,&nbsp;Nikhil Bommakanti ,&nbsp;Gina Yu ,&nbsp;Julie Rosenthal ,&nbsp;Yannis M. Paulus","doi":"10.1016/j.aopr.2025.10.003","DOIUrl":"10.1016/j.aopr.2025.10.003","url":null,"abstract":"<div><h3>Background</h3><div>Diabetic retinopathy (DR) is a sight-threatening retinal disease with pathological mediation by vascular endothelial growth factor (VEGF). Intravitreal anti-VEGF injections are commonly used to manage DR. Ultra-wide-field fluorescein angiography (UWF-FA) can assess DR severity and characterize the amount of non-perfused retina (non-perfusion area, NP), neovascularization (neovascular area, NV), and foveal avascular zone (FAZ). However, the association between anti-VEGF treatment and NP, NV, and FAZ characterized by UWF-FA is not well established.</div></div><div><h3>Methods</h3><div>A retrospective, single-center cohort study involved eyes of patients with Type 1 or 2 diabetes mellitus with at least one UWF-FA image. The area of the FAZ, NP area, and NV area of UWF-FA images was calculated. Stepwise multivariate logistic regression was used to identify patient and eye-level factors that were significant predictors of FAZ, NP, and NV. Causal model analyses estimated intravitreal treatment effects on FAZ, NP, and NV over time.</div></div><div><h3>Results</h3><div>The study included 705 eyes from patients with a mean (SD) age of 59.2 (13.2) years, and 56.3% were male. Eyes were treated with a mean (SD) of 5.6 (7.7) anti-VEGF and 0.63 (1.94) intravitreal steroid injections. Each incremental increase in Early Treatment Diabetic Retinopathy Study (ETDRS) severity of DR on initial presentation by clinical exam was associated with a 12.86 mm² increase in total NP and a 0.73 mm² increase in NV area. Anti-VEGF and steroid treatment had no significant impact on FAZ area. The presence of intravitreal steroid treatment was associated with an estimated decrease of −0.909 mm² in total NV area (<em>P</em> = 0.009). Each additional anti-VEGF injection decreased NV area by −0.239 mm² (<em>P</em> = 0.031). Any steroid use also led to an estimated −10.79 mm² decrease in NP (<em>P</em> = 0.041). Each additional anti-VEGF injection was predictive of a −2.54 mm² decrease in NP (<em>P</em> = 0.019). Each additional steroid injection was predictive of a −13.68 mm² decrease in NP (<em>P</em> = 0.001).</div></div><div><h3>Conclusions</h3><div>Intravitreal treatment was significantly associated with reduced NV and NP on UWF-FA. Intravitreal treatments were not predictive of FAZ changes. These findings suggest total retinal NV and NP areas may provide utility as UWF-FA biomarkers for assessing intravitreal treatment response.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 20-25"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841672","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}

{"title":"Instrument-based, non-cycloplegic versus cycloplegic refraction in pediatric and young adult populations (≤25 years): A systematic review and meta-analysis","authors":"Ana Roque ,&nbsp;Amélia Fernandes Nunes ,&nbsp;Henrique Nascimento ,&nbsp;Clara Martinez-Perez","doi":"10.1016/j.aopr.2025.11.004","DOIUrl":"10.1016/j.aopr.2025.11.004","url":null,"abstract":"<div><h3>Background</h3><div>Accurate refractive assessment in children and young adults is critical to prevent amblyopia and strabismus, conditions that may arise from uncorrected hyperopia. Although non-cycloplegic autorefractors and photoscreeners are increasingly used for vision screening due to their practicality and high testability, residual accommodation often introduces systematic measurement bias. The debate regarding the necessity of cycloplegia has intensified, particularly in large-scale epidemiological studies and screening programs, highlighting the need for an evidence-based synthesis.</div></div><div><h3>Methods</h3><div>This systematic review and meta-analysis, conducted according to PRISMA and AMSTAR-2 standards and registered in PROSPERO (CRD420251134665), synthesized data from 54 comparative studies, with 24 included in quantitative analyses.</div></div><div><h3>Results</h3><div>Compared with cycloplegic reference methods, non-cycloplegic autorefractors and photoscreeners consistently underestimated refractive error, showing pooled mean differences of −0.65 D (95% CI: −0.84 to −0.45; 95% PI: −1.50 to +0.20 D) and −0.78 D (95% CI: −1.12 to −0.44; 95% PI: −1.70 to +0.10 D), respectively. These prediction intervals illustrate the wide variability expected across future studies and populations. The bias was most pronounced in younger children and hyperopic eyes, reflecting the impact of accommodative tone. Despite device-specific differences, no method fully corrected this systematic error. Testability exceeded 95% across most devices, reinforcing their feasibility for population-level screening. However, the certainty of evidence was rated as low due to heterogeneity and observational design limitations.</div></div><div><h3>Conclusions</h3><div>Non-cycloplegic methods systematically underestimate hyperopia and therefore cannot replace cycloplegia for definitive diagnosis or spectacle prescription in pediatric populations. Cycloplegic assessment remains essential to detect amblyogenic refractive errors accurately. Non-cycloplegic methods may be integrated into large-scale screening programs for initial case detection, but positive or borderline cases must undergo cycloplegic confirmation to ensure safe and effective clinical management.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 56-67"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146078128","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}

{"title":"Comparative analysis of three toric intraocular lenses: clinical outcomes and rotational stability in patients with corneal astigmatism","authors":"Qiaomei Tang ,&nbsp;Wenxin Yang ,&nbsp;Yueyang Zhong ,&nbsp;Xiaohui Song ,&nbsp;Jianghua Hu ,&nbsp;Zhenwei Qin ,&nbsp;Yinhui Yu ,&nbsp;Yibo Yu","doi":"10.1016/j.aopr.2025.10.007","DOIUrl":"10.1016/j.aopr.2025.10.007","url":null,"abstract":"<div><h3>Pu<em>r</em>poses</h3><div>This study aims to evaluate the clinical outcomes and rotational stability of three different toric intraocular lenses (IOLs; Tecnis ZCT, Tecnis ZCU and AcrySof IQ) used in femtosecond laser-assisted cataract surgery (FLACS) and investigate potential variables that may impact the rotational stability.</div></div><div><h3>Methods</h3><div>This retrospective study involved 126 eyes from 111 patients who received FLACS with toric IOL implantation. Patients were allocated to three groups: one received the AcrySof IQ toric IOL, another the Tecnis ZCT, and the other the Tecnis ZCU. Ophthalmic parameters such as axial length (AL), anterior chamber depth (ACD), and white-to-white (WTW) distance, were determined preoperatively. Postoperative evaluations were conducted at 3 months.</div></div><div><h3>Results</h3><div>No significant differences in BCDVA or residual astigmatism (RAS) postoperatively were observed among the groups (<em>P</em> &gt; 0.05). No significant differences were found among the three toric IOLs in mean absolute rotation or directional distribution (<em>P</em> &gt; 0.05). In eyes with axial length ≥ 24.00 mm, the Tecnis ZCT toric IOL demonstrated significantly poorer rotational stability compared to the AcrySof IQ (Tecnis ZCT: 6.41 ± 5.23°, AcrySof IQ: 3.40 ± 2.95°, <em>P</em> = 0.032). Pairwise comparisons showed significant differences in capsule-IOL overlap between the AcrySof IQ and Tecnis ZCT groups (<em>P</em> = 0.014), as well as between the Tecnis ZCT and Tecnis ZCU groups (<em>P</em> &lt; 0.001), but not between the AcrySof IQ and Tecnis ZCU (<em>P</em> = 0.187). There was no significant difference (<em>P</em> &gt; 0.05) in IOL decentration among the three groups. In the Tecnis ZCT group, toric IOL rotation demonstrated a statistically significant association with the WTW distance (Pearson's <em>r</em> = 0.313, <em>P</em> = 0.034). For Tecnis ZCU, a significant correlation was observed between toric IOL rotation and the AL (Pearson's <em>r</em> = 0.325, <em>P</em> = 0.041),</div></div><div><h3>Conclusions</h3><div>The three toric IOLs showed comparable visual and refractive outcomes at 3 months. The Tecnis ZCT demonstrated inferior performance in rotational stability and postoperative alignment compared to the AcrySof IQ and Tecnis ZCU.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 42-49"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977946","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}

{"title":"Benchmarking large multimodal models for ophthalmic visual question answering with OphthalWeChat","authors":"Pusheng Xu ,&nbsp;Xia Gong ,&nbsp;Xiaolan Chen ,&nbsp;Weiyi Zhang ,&nbsp;Jiancheng Yang ,&nbsp;Bingjie Yan ,&nbsp;Meng Yuan ,&nbsp;Yalin Zheng ,&nbsp;Mingguang He ,&nbsp;Danli Shi","doi":"10.1016/j.aopr.2025.10.006","DOIUrl":"10.1016/j.aopr.2025.10.006","url":null,"abstract":"<div><h3>Purpose</h3><div>To develop a bilingual multimodal visual question answering (VQA) benchmark for evaluating Vision–language models (VLMs) in ophthalmology.</div></div><div><h3>Methods</h3><div>In this cross-sectional study, ophthalmic image posts and associated captions published between Jan 1, 2016, and Dec 31, 2024, were collected from WeChat Official Accounts. Based on these captions, bilingual question–answer (QA) pairs in Chinese and English were generated using GPT-4o-mini. QA pairs were categorized into six subsets by question type and language: binary (Binary_CN, Binary_EN), single-choice (Single-choice_CN, Single-choice_EN), and open-ended (Open-ended_CN, Open-ended_EN). The benchmark was used to evaluate six VLMs: GPT-4o, Gemini 2.0 Flash, Qwen2.5-VL-72B-Instruct, Janus-Pro-7B, InternVL3-8B, and HealthGPT-L14. Primary outcome was overall accuracy; secondary outcomes included subset-, subspeciality-, and modality-specific accuracy. Performance on open-ended questions were also quantified using language-based metrics, including AlignScore, BARTScore, BERTScore, BLEU, CIDEr, METEOR, and ROUGE_L. Error types in open-ended responses were manually analyzed through stratified sampling.</div></div><div><h3>Results</h3><div>OphthalWeChat included 3469 images and 30120 QA pairs cover 9 ophthalmic subspecialties, 548 conditions, 29 imaging modalities, and 68 modality combinations. Gemini 2.0 Flash achieved the highest overall accuracy (0.555), significantly outperforming GPT-4o (0.527), Qwen2.5-VL-72B-Instruct (0.520), HealthGPT-L14 (0.502), InternVL3-L14 (0.453), and Janus-Pro-7B (0.333) (all <em>P</em> &lt; 0.001). It also led in both Chinese (0.551) and English subsets (0.559). By subset, Gemini 2.0 Flash excelled in Binary_CN (0.687) and Single-choice_CN (0.666); HealthGPT-L14 performed best in Single-choice_EN (0.739); while GPT-4o ranked highest in Binary_EN (0.717), Open-ended_CN (0.254), and Open-ended_EN (0.271). Language-based metrics showed inconsistent rankings relative to accuracy in open-ended subsets. Performance varied across subspecialties and modalities, with Gemini 2.0 Flash leading in 6 of 9 subspecialties and 11 of top-15 imaging modalities. Error types analysis revealed lesion/diagnosis errors as the most frequent (35.6%–50.6%), followed by anatomical location errors (28.3%–37.5%).</div></div><div><h3>Conclusions</h3><div>This study presents the first bilingual VQA benchmark for ophthalmology, distinguished by its real-world context and inclusion of multiple examinations per patient. The dataset enables quantitative evaluation of VLMs, supporting the development of accurate and specialized AI systems for eye care.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 33-41"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977945","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}

{"title":"The influence of ocular parameters on intraocular pressure values measured by different devices and the development of an optimized prediction model in a myopic population","authors":"Wei Gan ,&nbsp;Yu Zhang ,&nbsp;Shizhao Yang ,&nbsp;Peiyun He ,&nbsp;Qingqing Hu ,&nbsp;Zhiyu Du","doi":"10.1016/j.aopr.2025.12.003","DOIUrl":"10.1016/j.aopr.2025.12.003","url":null,"abstract":"<div><h3>Objective</h3><div>To investigate the associations between ocular parameters—mean keratometry (Km), spherical equivalent (SE), central corneal thickness (CCT), axial length (AL), corneal hysteresis (CH), and corneal resistance factor (CRF)—and intraocular pressure (IOP) measurements obtained by non-contact tonometry (IOP_NCT), Schiøtz tonometry (IOP_Schiøtz), dynamic contour tonometry (IOP_DCT), corneal-compensated (IOP_cc), and Goldmann-correlated IOP (IOP_g) in a myopic population. Correlation and regression analyses were performed to identify the IOP measurement method least affected by ocular parameters. An optimized prediction model was then constructed based on conventional tonometry.</div></div><div><h3>Methods</h3><div>This study included 64 patients (128 eyes). All participants underwent comprehensive ocular examinations (CCT, AL, SE, Km, CH, and CRF) and IOP measurements using multiple tonometers. Pearson correlation and multiple linear regression analyses were employed to identify key influencers on IOP readings. An optimized prediction model was subsequently developed using NCT, CCT, CH, and CRF as independent variables, and its goodness-of-fit was evaluated.</div></div><div><h3>Results</h3><div>Age, AL, Km, IOP_Schiøtz, IOP_DCT, IOP_cc, and IOP_g showed statistically significant differences between sexes. Multiple regression identified CRF as the strongest positive predictor for all IOP measurements <em>(P</em> &lt; 0.001), while CH was a consistent negative predictor (<em>P</em> &lt; 0.01). Notably, dynamic contour tonometry (DCT) demonstrated the smallest regression coefficients for CH (B = −0.745) and CRF (B = 0.812), indicating minimal influence from corneal parameters. An optimized prediction model was developed: DCT-pred = 18.657 + 0.394 × NCT − 0.008 × CCT − 0.415 × CH + 0.107 × CRF (MSE = 3.30).</div></div><div><h3>Conclusions</h3><div>DCT measurements were less influenced by CCT, CH, and CRF, supporting their potential reliability for IOP assessment in patients with myopia. The optimized prediction model based on multiple regression analysis was able to predict DCT values to some extent, thereby reducing the influence of corneal biomechanical properties on IOP measurement. However, the model requires further validation and refinement in studies with larger sample sizes.</div></div>","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages 87-94"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146215041","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}

{"title":"TOC","authors":"","doi":"10.1016/S2667-3762(26)00023-5","DOIUrl":"10.1016/S2667-3762(26)00023-5","url":null,"abstract":"","PeriodicalId":72103,"journal":{"name":"Advances in ophthalmology practice and research","volume":"6 1","pages":"Pages iii-v"},"PeriodicalIF":3.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147449115","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}