{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T02:02:56Z","timestamp":1784772176693,"version":"3.55.0"},"reference-count":28,"publisher":"BMJ","issue":"8","content-domain":{"domain":["bmj.com"],"crossmark-restriction":true},"short-container-title":["Br J Ophthalmol"],"accepted":{"date-parts":[[2025,12,20]]},"published-print":{"date-parts":[[2026,8]]},"abstract":"<jats:sec>\n                    <jats:title>Background\/Aims<\/jats:title>\n                    <jats:p>To evaluate and compare the diagnostic capabilities of advanced large language models (LLMs) in interpreting ophthalmological fundus images across diverse pathologies.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We evaluated eight leading multimodal LLMs (GPT-4.5, Claude 3.7 Sonnet, Grok-2, Deepseek Cognition V2, Qwen2 72B, Gemini 2.0 Pro, Llama 3 405B and Mixtral 8\u00d722B) on their ability to interpret 100 fundus images representing various ophthalmological conditions. Performance was assessed using validated charts for diagnostic accuracy, specificity, sensitivity, consistency, relevance and explanation quality.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>GPT-4.5 achieved the highest overall diagnostic accuracy (65.0%), followed by Gemini 2.0 Pro (63.0%). All models showed varied performance across pathology categories, with rhegmatogenous pathologies being most accurately identified (Gemini 2.0 Pro: 81.3%, GPT-4.5: 75.0%) and myopic maculopathy (mean accuracy 21.8%) being particularly challenging. The remaining models performed significantly worse: Deepseek Cognition V2 (52.0%), Claude 3.7 Sonnet (52.0%), Qwen2 72B (49.0%), Llama 3 405B (48.0%), Grok-2 (47.0%) and Mixtral 8\u00d722B (46.0%). Lower-performing models frequently declined to provide diagnoses, with refusal rates from 8.0% (Claude 3.7 Sonnet) to 19.0% (Mixtral 8\u00d722B).<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>Current LLMs show promising but limited capabilities in ophthalmological image interpretation. While performance on common conditions like retinal detachments and age-related macular degeneration is moderately good, significant challenges remain with rare conditions, myopic pathologies and complex vascular disorders. The competitive performance between GPT-4.5 and Gemini 2.0 Pro, with each excelling in different pathology categories, suggests that leveraging their complementary strengths might offer improved diagnostic support.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1136\/bjo-2025-327634","type":"journal-article","created":{"date-parts":[[2026,1,14]],"date-time":"2026-01-14T18:20:20Z","timestamp":1768414820000},"page":"882-891","update-policy":"https:\/\/doi.org\/10.1136\/crossmarkpolicy","source":"Crossref","is-referenced-by-count":0,"title":["Advanced analysis of leading large language models for diagnostic accuracy in retinal imaging"],"prefix":"10.1136","volume":"110","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-2979-1638","authenticated-orcid":false,"given":"Matteo Mario","family":"Carl\u00e0","sequence":"first","affiliation":[{"name":"Ophthalmology Department, Fondation Adolphe de Rothschild, Paris, France"},{"name":"Ophthalmology Department, Catholic University of the Sacred Heart Faculty of Medicine and Surgery, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9996-9871","authenticated-orcid":false,"given":"Emanuele","family":"Crincoli","sequence":"additional","affiliation":[{"name":"Ophthalmology Department, Catholic University of the Sacred Heart Faculty of Medicine and Surgery, Rome, Italy"},{"name":"Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1551-8120","authenticated-orcid":false,"given":"Fiammetta","family":"Catania","sequence":"additional","affiliation":[{"name":"Department of Biomedical Science, Fondation Adolphe de Rothschild, Paris, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0004-3513-5826","authenticated-orcid":false,"given":"Laura","family":"De Luca","sequence":"additional","affiliation":[{"name":"Ophthalmology, University of Messina, Messina, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Federico","family":"Giannuzzi","sequence":"additional","affiliation":[{"name":"Ophthalmology Department, Catholic University of the Sacred Heart Faculty of Medicine and Surgery, Rome, Italy"},{"name":"Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Francesco","family":"Boselli","sequence":"additional","affiliation":[{"name":"Ophthalmology Department, Catholic University of the Sacred Heart Faculty of Medicine and Surgery, Rome, Italy"},{"name":"Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9500-049X","authenticated-orcid":false,"given":"Gloria","family":"Gambini","sequence":"additional","affiliation":[{"name":"Ophthalmology Department, Catholic University of the Sacred Heart Faculty of Medicine and Surgery, Rome, Italy"},{"name":"Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Carlos","family":"Mateo","sequence":"additional","affiliation":[{"name":"Ophthalmology Department, Instituto de Microcirugia Ocular, Barcelona, Barcelona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6302-063X","authenticated-orcid":false,"given":"Stanislao","family":"Rizzo","sequence":"additional","affiliation":[{"name":"Ophthalmology Department, Catholic University of the Sacred Heart Faculty of Medicine and Surgery, Rome, Italy"},{"name":"Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"239","published-online":{"date-parts":[[2026,1,14]]},"reference":[{"key":"2026072202101934000_110.8.882.1","doi-asserted-by":"publisher","DOI":"10.1136\/bjophthalmol-2018-313173"},{"key":"2026072202101934000_110.8.882.2","doi-asserted-by":"publisher","DOI":"10.1001\/jama.2016.17216"},{"key":"2026072202101934000_110.8.882.3","doi-asserted-by":"publisher","DOI":"10.1016\/j.ophtha.2021.03.003"},{"key":"2026072202101934000_110.8.882.4","first-page":"156","article-title":"PaLM 2 for Medical Question Answering: Advancing Medical AI with Advanced Language Models","volume":"6","author":"Huang","year":"2023","journal-title":"NPJ Digit Med"},{"key":"2026072202101934000_110.8.882.5","article-title":"Large language models in medicine: the potentials and pitfalls","volume":"30","author":"Thirunavukarasu","year":"2023","journal-title":"BMJ Health Care Inform"},{"key":"2026072202101934000_110.8.882.6","first-page":"2522","article-title":"Use of large language models in healthcare","volume":"29","author":"Barnett","year":"2023","journal-title":"Nat Med"},{"key":"2026072202101934000_110.8.882.7","first-page":"160","article-title":"GPT-4V for medical image understanding: A preliminary evaluation and potential implications","volume":"6","author":"Tu","year":"2023","journal-title":"NPJ Digit Med"},{"key":"2026072202101934000_110.8.882.8","first-page":"158","article-title":"Capabilities of GPT-4 on medical challenge problems","volume":"6","author":"Nori","year":"2023","journal-title":"NPJ Digit Med"},{"key":"2026072202101934000_110.8.882.9","doi-asserted-by":"crossref","first-page":"259","DOI":"10.1038\/s41586-023-05881-4","article-title":"Foundation models for generalist medical artificial intelligence","volume":"616","author":"Moor","year":"2023","journal-title":"Nature New Biol"},{"key":"2026072202101934000_110.8.882.10","first-page":"2198","article-title":"Evaluating medical chatbots during high-stakes medical emergencies","volume":"29","author":"Agrawal","year":"2023","journal-title":"Nat Med"},{"key":"2026072202101934000_110.8.882.11","doi-asserted-by":"crossref","unstructured":"Bossuyt PM , Reitsma JB , Bruns DE , et al . STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 2015;351. doi:10.1136\/bmj.h5527","DOI":"10.1136\/bmj.h5527"},{"key":"2026072202101934000_110.8.882.12","doi-asserted-by":"publisher","DOI":"10.1016\/j.jclinepi.2010.03.002"},{"key":"2026072202101934000_110.8.882.13","unstructured":"Vagias WM . Likert-type scale response anchors. Clemson International Institute for Tourism & Research Development 2006."},{"key":"2026072202101934000_110.8.882.14","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-020-1034-x"},{"key":"2026072202101934000_110.8.882.15","doi-asserted-by":"crossref","unstructured":"Collins GS , Reitsma JB , Altman DG , et al . Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ 2015;350. doi:10.1136\/bmj.g7594","DOI":"10.1136\/bmj.g7594"},{"key":"2026072202101934000_110.8.882.16","doi-asserted-by":"publisher","DOI":"10.1136\/bjo-2023-325143"},{"key":"2026072202101934000_110.8.882.17","doi-asserted-by":"crossref","first-page":"2945","DOI":"10.1007\/s00417-024-06470-5","article-title":"Large language models as assistance for glaucoma surgical cases: a ChatGPT vs. Google Gemini comparison","volume":"262","author":"Carl\u00e0","year":"2024","journal-title":"Graefes Arch Clin Exp Ophthalmol"},{"key":"2026072202101934000_110.8.882.18","first-page":"1967","article-title":"Retinal Imaging Analysis Performed By ChatGPT-4o And Gemini Advanced","volume":"44","author":"Carl\u00e0","year":"2024","journal-title":"The Turning Point Of The Revolution? Retina"},{"key":"2026072202101934000_110.8.882.19","first-page":"895","article-title":"Artificial intelligence with ChatGPT 4: a large language model in support of ocular oncology cases","volume":"45","author":"Giannuzzi","year":"2024","journal-title":"Int Ophthalmol"},{"key":"2026072202101934000_110.8.882.20","first-page":"527","article-title":"Gemini AI vs. ChatGPT: A comprehensive examination alongside ophthalmology residents in medical knowledge","volume":"263","author":"Bahir","year":"2025","journal-title":"Graefes Arch Clin Exp Ophthalmol"},{"key":"2026072202101934000_110.8.882.21","doi-asserted-by":"crossref","first-page":"3223","DOI":"10.7150\/thno.100786","article-title":"Artificial intelligence-enhanced retinal imaging as a biomarker for systemic diseases","volume":"15","author":"Wang","year":"2025","journal-title":"Theranostics"},{"key":"2026072202101934000_110.8.882.22","doi-asserted-by":"crossref","first-page":"445","DOI":"10.1016\/j.ajo.2024.12.021","article-title":"Ophthalmology Journals\u2019 Guidelines on Generative Artificial Intelligence: A Comprehensive Analysis","volume":"271","author":"Huang","year":"2025","journal-title":"Am J Ophthalmol"},{"key":"2026072202101934000_110.8.882.23","doi-asserted-by":"crossref","first-page":"785","DOI":"10.1038\/s41433-024-03488-1","article-title":"The use of artificial intelligence based chat bots in ophthalmology triage","volume":"39","author":"David","year":"2025","journal-title":"Eye (Lond)"},{"key":"2026072202101934000_110.8.882.24","doi-asserted-by":"crossref","first-page":"100063","DOI":"10.1016\/j.ajoint.2024.100063","article-title":"Testing the power of Google DeepMind: Gemini versus ChatGPT 4 facing a European ophthalmology examination","volume":"1","author":"Carl\u00e0","year":"2024","journal-title":"AJO International"},{"key":"2026072202101934000_110.8.882.25","article-title":"Physicians\u2019 Perspectives on ChatGPT in Ophthalmology: Insights on Artificial Intelligence (AI) Integration in Clinical Practice","volume":"17","author":"Ahmed","year":"2025","journal-title":"Cureus"},{"key":"2026072202101934000_110.8.882.26","doi-asserted-by":"crossref","first-page":"27","DOI":"10.1167\/tvst.14.1.27","article-title":"Artificial Intelligence for Optical Coherence Tomography in Glaucoma","volume":"14","author":"Djulbegovic","year":"2025","journal-title":"Transl Vis Sci Technol"},{"key":"2026072202101934000_110.8.882.27","doi-asserted-by":"crossref","first-page":"1235","DOI":"10.1038\/s41433-025-03746-w","article-title":"New frontiers in generative artificial intelligence - enhancing custom GPTs with ophthalmology guidelines","volume":"39","author":"Taloni","year":"2025","journal-title":"Eye (Lond)"},{"key":"2026072202101934000_110.8.882.28","doi-asserted-by":"crossref","first-page":"64","DOI":"10.1038\/s41746-025-01461-0","article-title":"Multimodal machine learning enables AI chatbot to diagnose ophthalmic diseases and provide high-quality medical responses","volume":"8","author":"Ma","year":"2025","journal-title":"NPJ Digit Med"}],"container-title":["British Journal of Ophthalmology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1136\/bjo-2025-327634","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T01:47:14Z","timestamp":1784771234000},"score":1,"resource":{"primary":{"URL":"https:\/\/bjo.bmj.com\/lookup\/doi\/10.1136\/bjo-2025-327634"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,14]]},"references-count":28,"journal-issue":{"issue":"8","published-online":{"date-parts":[[2026,7,22]]},"published-print":{"date-parts":[[2026,8]]}},"alternative-id":["10.1136\/bjo-2025-327634"],"URL":"https:\/\/doi.org\/10.1136\/bjo-2025-327634","relation":{},"ISSN":["0007-1161","1468-2079"],"issn-type":[{"value":"0007-1161","type":"print"},{"value":"1468-2079","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,14]]}}}