Can Multimodal Large Language Models Diagnose Diabetic Retinopathy from Fundus Photos? A Quantitative Evaluation

To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features. A single-center retrospective study. Patients diagnosed with prediabetes and diabetes. Ultra-widefield fundus images from pati...

Full description

Saved in:
Bibliographic Details
Published in:Ophthalmology science (Online) Vol. 6; no. 1; p. 100911
Main Authors: Most, Jesse A., Walker, Evan H., Mehta, Nehal N., Nagel, Ines D., Chen, Jimmy S., Russell, Jonathan F., Scott, Nathan L., Borooah, Shyamanga
Format: Journal Article
Language:English
Published: Netherlands Elsevier Inc 01.01.2026
Elsevier
Subjects:
ISSN:2666-9145, 2666-9145
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features. A single-center retrospective study. Patients diagnosed with prediabetes and diabetes. Ultra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity. Accuracy, sensitivity, and specificity of diagnosis. A total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7–60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher (P < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608–0.566) and sensitivity (0.618–0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870–0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866. Multimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
AbstractList Objective: To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features. Design: A single-center retrospective study. Subjects: Patients diagnosed with prediabetes and diabetes. Methods: Ultra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity. Main Outcome Measures: Accuracy, sensitivity, and specificity of diagnosis. Results: A total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7–60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher (P < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608–0.566) and sensitivity (0.618–0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870–0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866. Conclusions: Multimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility. Financial Disclosure(s): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features. A single-center retrospective study. Patients diagnosed with prediabetes and diabetes. Ultra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity. Accuracy, sensitivity, and specificity of diagnosis. A total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7–60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher (P < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608–0.566) and sensitivity (0.618–0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870–0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866. Multimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features. A single-center retrospective study. Patients diagnosed with prediabetes and diabetes. Ultra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity. Accuracy, sensitivity, and specificity of diagnosis. A total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7-60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher ( < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608-0.566) and sensitivity (0.618-0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870-0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866. Multimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features.ObjectiveTo evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features.A single-center retrospective study.DesignA single-center retrospective study.Patients diagnosed with prediabetes and diabetes.SubjectsPatients diagnosed with prediabetes and diabetes.Ultra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity.MethodsUltra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity.Accuracy, sensitivity, and specificity of diagnosis.Main Outcome MeasuresAccuracy, sensitivity, and specificity of diagnosis.A total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7-60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher (P < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608-0.566) and sensitivity (0.618-0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870-0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866.ResultsA total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7-60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher (P < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608-0.566) and sensitivity (0.618-0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870-0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866.Multimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility.ConclusionsMultimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility.Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.Financial DisclosuresProprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
ObjectiveTo evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image analysis features. DesignA single-center retrospective study. SubjectsPatients diagnosed with prediabetes and diabetes. MethodsUltra-widefield fundus images from patients seen at the University of California, San Diego, were graded for DR severity by 3 retina specialists using the ETDRS classification system to establish ground truth. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, and Perplexity Llama 3.1 Sonar/Default) were tested using 4 distinct prompts. These assessed multiple-choice disease diagnosis, binary disease classification, and disease severity. Multimodal large language models were assessed for accuracy, sensitivity, and specificity in identifying the presence or absence of DR and relative disease severity. Main Outcome MeasuresAccuracy, sensitivity, and specificity of diagnosis. ResultsA total of 309 eyes from 188 patients were included in the study. The average patient age was 58.7 (56.7–60.7) years, with 55.3% being female. After specialist grading, 70.2% of eyes had DR of varying severity, and 29.8% had no DR. For disease identification with multiple choices provided, Claude and ChatGPT scored significantly higher ( P < 0.0006, per Bonferroni correction) than other MLLMs for accuracy (0.608–0.566) and sensitivity (0.618–0.641). In binary DR versus no DR classification, accuracy was the highest for ChatGPT (0.644) and Perplexity (0.602). Sensitivity varied (ChatGPT [0.539], Perplexity [0.488], Claude [0.179], and Gemini [0.042]), whereas specificity for all models was relatively high (range: 0.870–0.989). For the DR severity prompt with the best overall results (Prompt 3.1), no significant differences between models were found in accuracy (Perplexity [0.411], ChatGPT [0.395], Gemini [0.392], and Claude [0.314]). All models demonstrated low sensitivity (Perplexity [0.247], ChatGPT [0.229], Gemini [0.224], and Claude [0.184]). Specificity ranged from 0.840 to 0.866. ConclusionsMultimodal large language models are powerful tools that may eventually assist retinal image analysis. Currently, however, there is variability in the accuracy of image analysis, and diagnostic performance falls short of clinical standards for safe implementation in DR diagnosis and grading. Further training and optimization of common errors may enhance their clinical utility. Financial Disclosure(s)Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
ArticleNumber 100911
Author Scott, Nathan L.
Walker, Evan H.
Borooah, Shyamanga
Most, Jesse A.
Chen, Jimmy S.
Nagel, Ines D.
Mehta, Nehal N.
Russell, Jonathan F.
Author_xml – sequence: 1
  givenname: Jesse A.
  orcidid: 0000-0002-2582-4503
  surname: Most
  fullname: Most, Jesse A.
  organization: Jacobs Retina Center, Shiley Eye Institute, University of California San Diego, La Jolla, California
– sequence: 2
  givenname: Evan H.
  surname: Walker
  fullname: Walker, Evan H.
  organization: Viterbi Family Department of Ophthalmology and Shiley Eye Institute, University of California San Diego, La Jolla, California
– sequence: 3
  givenname: Nehal N.
  surname: Mehta
  fullname: Mehta, Nehal N.
  organization: Jacobs Retina Center, Shiley Eye Institute, University of California San Diego, La Jolla, California
– sequence: 4
  givenname: Ines D.
  surname: Nagel
  fullname: Nagel, Ines D.
  organization: Jacobs Retina Center, Shiley Eye Institute, University of California San Diego, La Jolla, California
– sequence: 5
  givenname: Jimmy S.
  surname: Chen
  fullname: Chen, Jimmy S.
  organization: Viterbi Family Department of Ophthalmology and Shiley Eye Institute, University of California San Diego, La Jolla, California
– sequence: 6
  givenname: Jonathan F.
  surname: Russell
  fullname: Russell, Jonathan F.
  organization: Department of Ophthalmology and Visual Sciences, University of Iowa Hospitals and Clinics, Iowa City, Iowa
– sequence: 7
  givenname: Nathan L.
  surname: Scott
  fullname: Scott, Nathan L.
  organization: Jacobs Retina Center, Shiley Eye Institute, University of California San Diego, La Jolla, California
– sequence: 8
  givenname: Shyamanga
  surname: Borooah
  fullname: Borooah, Shyamanga
  email: sborooah@health.ucsd.edu
  organization: Jacobs Retina Center, Shiley Eye Institute, University of California San Diego, La Jolla, California
BackLink https://www.ncbi.nlm.nih.gov/pubmed/41030829$$D View this record in MEDLINE/PubMed
BookMark eNqFUk1v1DAQjVAR_aB_gAPKkcsuthN7Y4SoqqWFSlvxLXGzJs546yVrBztZdf89DikVQgIu46fRe280fnOcHTjvMMueUDKnhIrnm_mt7-KcEcZTg0hKH2RHTAgxk7TkB7_hw-w0xg0hiUkLVtJH2WFJSUEqJo-ybgkuvx7a3m59A22-grDGVN16gASufYNtzF9bWDsfcQQ19lbnH1N1voP-Zp-b4Lf55eCaIebvb3zv41l-nn8YwPW2h97uML_YQTsk6N3j7KGBNuLp3XuSfbm8-Lx8O1u9e3O1PF_NNGe8n2kooCgaWYCkErkpRWNqU1YVEt3QQksjSkm0FEZiuRgRmNIA8NTQgtbFSXY1-TYeNqoLdgthrzxY9bPhw1pBSJu0qDiQBaKpKmKg5LyukIlKEE0XpgRdmOT1bPLqgv8-YOzV1kaNbQsO_RBVwbigYwgkUZ_eUYd6i8394F8_nghsIujgYwxo7imUqDFZtVFjsmpMVk3JJtHLSZTCwJ3FoKK26DQ2NqDu00r23_JXf8h1a53V0H7DPcaNH4JLWSiqIlNEfRpPZ7wcxtPREPk1Gbz4u8H_pv8A4T7Ulg
Cites_doi 10.2337/dci23-0032
10.1177/1932296820967011
10.1159/000486284
10.2196/50638
10.1016/j.ajo.2024.02.012
10.1016/j.ophtha.2016.11.014
10.5888/pcd13.160193
10.4103/tjo.TJO-D-23-00193
10.4239/wjd.v4.i6.290
10.1111/nan.12997
10.1016/j.ophtha.2018.01.034
10.1038/s41574-020-00451-4
10.1186/s40942-024-00595-9
10.1038/s41433-024-03074-5
10.1016/j.xops.2024.100667
10.1016/j.ophtha.2013.05.004
10.1136/bjo-2023-324533
10.4172/2155-9570.1000582
10.1016/j.ophtha.2018.04.007
10.1016/j.diabres.2021.109119
10.1016/S2589-7500(20)30060-1
10.1016/j.ajo.2012.03.019
10.1016/j.ophtha.2019.09.025
10.1136/bjo-2024-325459
10.1097/APO.0000000000000403
10.2196/45312
10.1016/j.ajoint.2025.100111
10.1002/hsr2.2004
10.1016/j.xops.2024.100495
10.1136/bjo-2023-325054
10.3389/fmed.2025.1519768
10.1186/s40942-025-00670-9
10.1016/j.eclinm.2023.102387
10.1186/s40942-023-00511-7
10.2147/OPTH.S435052
10.1016/j.xops.2024.100600
10.1371/journal.pone.0198979
10.1016/j.compbiomed.2024.108709
ContentType Journal Article
Copyright 2025 American Academy of Ophthalmology
American Academy of Ophthalmology
2025 by the American Academy of Ophthalmologyé.
Copyright_xml – notice: 2025 American Academy of Ophthalmology
– notice: American Academy of Ophthalmology
– notice: 2025 by the American Academy of Ophthalmologyé.
DBID 6I.
AAFTH
AAYXX
CITATION
NPM
7X8
DOA
DOI 10.1016/j.xops.2025.100911
DatabaseName ScienceDirect Open Access Titles
Elsevier:ScienceDirect:Open Access
CrossRef
PubMed
MEDLINE - Academic
DOAJ Directory of Open Access Journals
DatabaseTitle CrossRef
PubMed
MEDLINE - Academic
DatabaseTitleList


PubMed
MEDLINE - Academic

Database_xml – sequence: 1
  dbid: DOA
  name: DOAJ Directory of Open Access Journals
  url: https://www.doaj.org/
  sourceTypes: Open Website
– sequence: 2
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 3
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Medicine
EISSN 2666-9145
EndPage 100911
ExternalDocumentID oai_doaj_org_article_5a07eef880fa455b8e26860c17f4ac3f
41030829
10_1016_j_xops_2025_100911
S266691452500209X
1_s2_0_S266691452500209X
Genre Journal Article
GroupedDBID .1-
.FO
0R~
AAEDW
AALRI
AAXUO
AAYWO
ACVFH
ADCNI
ADVLN
AEUPX
AFJKZ
AFPUW
AFRHN
AIGII
AITUG
AJUYK
AKBMS
AKRWK
AKYEP
ALMA_UNASSIGNED_HOLDINGS
AMRAJ
APXCP
EBS
FDB
GROUPED_DOAJ
M~E
OK1
ROL
RPM
Z5R
6I.
AAFTH
AAYXX
CITATION
NPM
7X8
ID FETCH-LOGICAL-c525t-ca3a33d93a919e5f46dfbf488e0cd13c9f6490c96f9e4790c9af4faa56f9c61b3
IEDL.DBID DOA
ISICitedReferencesCount 0
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001576792300002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 2666-9145
IngestDate Fri Oct 03 12:19:12 EDT 2025
Mon Oct 06 16:26:25 EDT 2025
Sat Oct 25 11:10:42 EDT 2025
Thu Nov 20 00:45:13 EST 2025
Sat Oct 25 17:13:26 EDT 2025
Sun Oct 19 01:20:41 EDT 2025
Sat Oct 25 11:12:41 EDT 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 1
Keywords Diabetic retinopathy
MLLM
UWF
AI
LLM
DR
Image analysis
NPDR
MC
Ultra-widefield fundus photography
Multimodal large language model
Artificial intelligence
PDR
ultra-widefield
nonproliferative diabetic retinopathy
large language model
multiple-choice
proliferative diabetic retinopathy
Language English
License This is an open access article under the CC BY-NC-ND license.
2025 by the American Academy of Ophthalmologyé.
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c525t-ca3a33d93a919e5f46dfbf488e0cd13c9f6490c96f9e4790c9af4faa56f9c61b3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ORCID 0000-0002-2582-4503
OpenAccessLink https://doaj.org/article/5a07eef880fa455b8e26860c17f4ac3f
PMID 41030829
PQID 3256110090
PQPubID 23479
PageCount 1
ParticipantIDs doaj_primary_oai_doaj_org_article_5a07eef880fa455b8e26860c17f4ac3f
proquest_miscellaneous_3256110090
pubmed_primary_41030829
crossref_primary_10_1016_j_xops_2025_100911
elsevier_sciencedirect_doi_10_1016_j_xops_2025_100911
elsevier_clinicalkeyesjournals_1_s2_0_S266691452500209X
elsevier_clinicalkey_doi_10_1016_j_xops_2025_100911
PublicationCentury 2000
PublicationDate 2026-01-01
PublicationDateYYYYMMDD 2026-01-01
PublicationDate_xml – month: 01
  year: 2026
  text: 2026-01-01
  day: 01
PublicationDecade 2020
PublicationPlace Netherlands
PublicationPlace_xml – name: Netherlands
PublicationTitle Ophthalmology science (Online)
PublicationTitleAlternate Ophthalmol Sci
PublicationYear 2026
Publisher Elsevier Inc
Elsevier
Publisher_xml – name: Elsevier Inc
– name: Elsevier
References Flaxel, Adelman, Bailey (bib25) 2020; 127
Sun, Saeedi, Karuranga (bib40) 2022; 183
Chia, Antaki, Zhou (bib15) 2024; 108
Mohammadi, Nguyen (bib9) 2024; 4
Agnihotri, Nagel, Artiaga (bib11) 2024; 5
Ruamviboonsuk, Chantra, Seresirikachorn (bib35) 2021; 10
Guo, Wei, Sun (bib17) 2024; 178
Raghu, S, S Devishamani (bib43) 2023; 17
Chen, Reddy, Al-Sharif (bib12) 2025; 5
Gilson, Safranek, Huang (bib8) 2023; 9
Xu, Chen, Zhao, Shi (bib20) 2024; 108
Hossain, Al-Mamun, Islam (bib1) 2024; 7
Antonetti, Silva, Stitt (bib2) 2021; 17
Wu, Nishida, Moghimi, Weinreb (bib22) 2024; 14
Cheong, Zhang, Tan (bib41) 2024; 108
Hansen, Tang (bib33) 2016; 7
Rajesh, Davidson, Lee, Lee (bib30) 2023; 46
Meskó (bib10) 2023; 25
Gupta, Al-Kazwini (bib44) 2024; 16
Tahir, Ullah, Tahir (bib28) 2025; 12
Mihalache, Huang, Cruz-Pimentel (bib16) 2024; 38
Joseph, Selvaraj, Mani (bib7) 2024; 263
Hu, Joseph, Li (bib29) 2024; 67
Nørgaard, Grauslund (bib4) 2018; 60
Ono, Dickson, Koga (bib32) 2024; 50
Wu, Fernandez-Loaiza, Sauma (bib26) 2013; 4
Krause, Gulshan, Rahimy (bib31) 2018; 125
Tufail, Rudisill, Egan (bib38) 2017; 124
Alqahtani, Alshareef, Aljadani (bib27) 2025; 11
Wong, Sun, Kawasaki (bib3) 2018; 125
Rizvi, Rizvi, Lalakia (bib34) 2023; 15
Lu (bib5) 2016; 13
bib18
Ferro, Roth, Zinkernagel, Anguita (bib42) 2023; 9
Piyasena, Murthy, Yip (bib6) 2019; 14
Xie, Nguyen, Hamzah (bib37) 2020; 2
bib39
Bellanda, Santos, Ferraz (bib13) 2024; 10
Silva, Cavallerano, Sun (bib23) 2012; 154
Fuller, Hu, Liu (bib36) 2020; 16
Aftab, Khan, VanderBeek (bib21) 2025; 2
Silva, Cavallerano, Sun (bib24) 2013; 120
Jalili, Jiravarnsirikul, Bowd (bib14) 2025; 5
Piyasena (10.1016/j.xops.2025.100911_bib6) 2019; 14
Gilson (10.1016/j.xops.2025.100911_bib8) 2023; 9
Guo (10.1016/j.xops.2025.100911_bib17) 2024; 178
Alqahtani (10.1016/j.xops.2025.100911_bib27) 2025; 11
Jalili (10.1016/j.xops.2025.100911_bib14) 2025; 5
Chen (10.1016/j.xops.2025.100911_bib12) 2025; 5
Tahir (10.1016/j.xops.2025.100911_bib28) 2025; 12
Krause (10.1016/j.xops.2025.100911_bib31) 2018; 125
Cheong (10.1016/j.xops.2025.100911_bib41) 2024; 108
Nørgaard (10.1016/j.xops.2025.100911_bib4) 2018; 60
Bellanda (10.1016/j.xops.2025.100911_bib13) 2024; 10
Ferro (10.1016/j.xops.2025.100911_bib42) 2023; 9
Raghu (10.1016/j.xops.2025.100911_bib43) 2023; 17
Wu (10.1016/j.xops.2025.100911_bib26) 2013; 4
Gupta (10.1016/j.xops.2025.100911_bib44) 2024; 16
Lu (10.1016/j.xops.2025.100911_bib5) 2016; 13
Rajesh (10.1016/j.xops.2025.100911_bib30) 2023; 46
Silva (10.1016/j.xops.2025.100911_bib24) 2013; 120
Flaxel (10.1016/j.xops.2025.100911_bib25) 2020; 127
Fuller (10.1016/j.xops.2025.100911_bib36) 2020; 16
Mihalache (10.1016/j.xops.2025.100911_bib16) 2024; 38
Xu (10.1016/j.xops.2025.100911_bib20) 2024; 108
Wu (10.1016/j.xops.2025.100911_bib22) 2024; 14
Ruamviboonsuk (10.1016/j.xops.2025.100911_bib35) 2021; 10
Hossain (10.1016/j.xops.2025.100911_bib1) 2024; 7
Mohammadi (10.1016/j.xops.2025.100911_bib9) 2024; 4
Meskó (10.1016/j.xops.2025.100911_bib10) 2023; 25
Ono (10.1016/j.xops.2025.100911_bib32) 2024; 50
Silva (10.1016/j.xops.2025.100911_bib23) 2012; 154
Chia (10.1016/j.xops.2025.100911_bib15) 2024; 108
Aftab (10.1016/j.xops.2025.100911_bib21) 2025; 2
Sun (10.1016/j.xops.2025.100911_bib40) 2022; 183
Rizvi (10.1016/j.xops.2025.100911_bib34) 2023; 15
Agnihotri (10.1016/j.xops.2025.100911_bib11) 2024; 5
Hansen (10.1016/j.xops.2025.100911_bib33) 2016; 7
Tufail (10.1016/j.xops.2025.100911_bib38) 2017; 124
Joseph (10.1016/j.xops.2025.100911_bib7) 2024; 263
Xie (10.1016/j.xops.2025.100911_bib37) 2020; 2
Antonetti (10.1016/j.xops.2025.100911_bib2) 2021; 17
Hu (10.1016/j.xops.2025.100911_bib29) 2024; 67
Wong (10.1016/j.xops.2025.100911_bib3) 2018; 125
References_xml – volume: 5
  year: 2025
  ident: bib14
  article-title: Glaucoma detection and feature identification via GPT-4V fundus image analysis
  publication-title: Ophthalmol Sci
– volume: 4
  start-page: 290
  year: 2013
  end-page: 294
  ident: bib26
  article-title: Classification of diabetic retinopathy and diabetic macular edema
  publication-title: World J Diabetes
– volume: 9
  year: 2023
  ident: bib8
  article-title: How does ChatGPT perform on the united States Medical Licensing Examination (USMLE)? The implications of large Language models for medical education and knowledge assessment
  publication-title: JMIR Med Educ
– volume: 17
  start-page: 4021
  year: 2023
  end-page: 4031
  ident: bib43
  article-title: The utility of ChatGPT in diabetic retinopathy risk assessment: a comparative study with clinical diagnosis
  publication-title: Clin Ophthalmol Auckl NZ
– volume: 120
  start-page: 2587
  year: 2013
  end-page: 2595
  ident: bib24
  article-title: Peripheral lesions identified by mydriatic ultrawide field imaging: distribution and potential impact on diabetic retinopathy severity
  publication-title: Ophthalmology
– volume: 46
  start-page: 1728
  year: 2023
  end-page: 1739
  ident: bib30
  article-title: Artificial intelligence and diabetic retinopathy: AI framework, prospective studies, head-to-head validation, and cost-effectiveness
  publication-title: Diabetes Care
– volume: 16
  start-page: 415
  year: 2020
  end-page: 427
  ident: bib36
  article-title: Five-year cost-effectiveness modeling of primary care-based, nonmydriatic automated retinal image analysis screening among low-income patients with diabetes
  publication-title: J Diabetes Sci Technol
– ident: bib39
  article-title: ChatGPT pricing
– volume: 127
  start-page: P66
  year: 2020
  end-page: P145
  ident: bib25
  article-title: Diabetic retinopathy preferred practice pattern
  publication-title: Ophthalmology
– volume: 7
  start-page: 4
  year: 2016
  ident: bib33
  article-title: Automated detection of diabetic retinopathy in three European populations - university of surrey
  publication-title: J Clin Exp Ophthalmol
– volume: 108
  year: 2024
  ident: bib15
  article-title: Foundation models in ophthalmology
  publication-title: Br J Ophthalmol
– volume: 125
  start-page: 1264
  year: 2018
  end-page: 1272
  ident: bib31
  article-title: Grader variability and the importance of reference standards for evaluating machine learning models for diabetic retinopathy
  publication-title: Ophthalmology
– volume: 7
  year: 2024
  ident: bib1
  article-title: Diabetes mellitus, the fastest growing global public health concern: early detection should be focused
  publication-title: Health Sci Rep
– ident: bib18
  article-title: GPT-4V(ision) system card
– volume: 5
  year: 2025
  ident: bib12
  article-title: Analysis of ChatGPT responses to ophthalmic cases: can ChatGPT think like an ophthalmologist?
  publication-title: Ophthalmol Sci
– volume: 125
  start-page: 1608
  year: 2018
  end-page: 1622
  ident: bib3
  article-title: Guidelines on diabetic eye care: the international council of ophthalmology recommendations for screening, follow-up, referral, and treatment based on resource settings
  publication-title: Ophthalmology
– volume: 10
  start-page: 307
  year: 2021
  end-page: 316
  ident: bib35
  article-title: Economic evaluations of artificial intelligence in ophthalmology
  publication-title: Asia Pac J Ophthalmol
– volume: 38
  start-page: 2491
  year: 2024
  end-page: 2493
  ident: bib16
  article-title: Artificial intelligence chatbot interpretation of ophthalmic multimodal imaging cases
  publication-title: Eye
– volume: 67
  year: 2024
  ident: bib29
  article-title: Population impact and cost-effectiveness of artificial intelligence-based diabetic retinopathy screening in people living with diabetes in Australia: a cost effectiveness analysis
  publication-title: EClinicalMedicine
– volume: 263
  start-page: 214
  year: 2024
  end-page: 230
  ident: bib7
  article-title: Diagnostic accuracy of artificial intelligence-based automated diabetic retinopathy screening in real-world settings: a systematic review and meta-analysis
  publication-title: Am J Ophthalmol
– volume: 124
  start-page: 343
  year: 2017
  end-page: 351
  ident: bib38
  article-title: Automated diabetic retinopathy image assessment software: diagnostic accuracy and cost-effectiveness compared with human graders
  publication-title: Ophthalmology
– volume: 108
  start-page: 1384
  year: 2024
  end-page: 1389
  ident: bib20
  article-title: Unveiling the clinical incapabilities: a benchmarking study of GPT-4V(ision) for ophthalmic multimodal image analysis
  publication-title: Br J Ophthalmol
– volume: 10
  start-page: 79
  year: 2024
  ident: bib13
  article-title: Applications of ChatGPT in the diagnosis, management, education, and research of retinal diseases: a scoping review
  publication-title: Int J Retina Vitr
– volume: 11
  start-page: 48
  year: 2025
  ident: bib27
  article-title: The efficacy of artificial intelligence in diabetic retinopathy screening: a systematic review and meta-analysis
  publication-title: Int J Retina Vitr
– volume: 108
  start-page: 1443
  year: 2024
  end-page: 1449
  ident: bib41
  article-title: Comparing generative and retrieval-based chatbots in answering patient questions regarding age-related macular degeneration and diabetic retinopathy
  publication-title: Br J Ophthalmol
– volume: 5
  year: 2024
  ident: bib11
  article-title: Large language models in ophthalmology: a review of publications from top ophthalmology journals
  publication-title: Ophthalmol Sci
– volume: 60
  start-page: 9
  year: 2018
  end-page: 17
  ident: bib4
  article-title: Automated screening for diabetic retinopathy – a systematic review
  publication-title: Ophthalmic Res
– volume: 14
  year: 2019
  ident: bib6
  article-title: Systematic review on barriers and enablers for access to diabetic retinopathy screening services in different income settings
  publication-title: PLoS One
– volume: 13
  start-page: E140
  year: 2016
  ident: bib5
  article-title: Divergent perceptions of barriers to diabetic retinopathy screening among patients and care providers, Los Angeles, California, 2014–2015
  publication-title: Prev Chronic Dis
– volume: 2
  start-page: e240
  year: 2020
  end-page: e249
  ident: bib37
  article-title: Artificial intelligence for teleophthalmology-based diabetic retinopathy screening in a national programme: an economic analysis modelling study
  publication-title: Lancet Digit Health
– volume: 25
  year: 2023
  ident: bib10
  article-title: Prompt engineering as an important emerging skill for medical professionals: tutorial
  publication-title: J Med Internet Res
– volume: 17
  start-page: 195
  year: 2021
  end-page: 206
  ident: bib2
  article-title: Current understanding of the molecular and cellular pathology of diabetic retinopathy
  publication-title: Nat Rev Endocrinol
– volume: 12
  year: 2025
  ident: bib28
  article-title: Artificial intelligence versus manual screening for the detection of diabetic retinopathy: a comparative systematic review and meta-analysis
  publication-title: Front Med
– volume: 15
  year: 2023
  ident: bib34
  article-title: Is artificial intelligence the cost-saving lens to diabetic retinopathy screening in Low- and middle-income countries?
  publication-title: Cureus
– volume: 14
  start-page: 454
  year: 2024
  end-page: 457
  ident: bib22
  article-title: Effects of prompt engineering on large language model performance in response to questions on common ophthalmic conditions
  publication-title: Taiwan J Ophthalmol
– volume: 9
  start-page: 71
  year: 2023
  ident: bib42
  article-title: Application and accuracy of artificial intelligence-derived large language models in patients with age related macular degeneration
  publication-title: Int J Retina Vitr
– volume: 16
  year: 2024
  ident: bib44
  article-title: Evaluating chatgpt's diagnostic accuracy in detecting fundus images
  publication-title: Cureus
– volume: 2
  year: 2025
  ident: bib21
  article-title: Evaluation of ChatGPT-4 in detecting referable diabetic retinopathy using single fundus images
  publication-title: AJO Int
– volume: 50
  year: 2024
  ident: bib32
  article-title: Evaluating the efficacy of few-shot learning for GPT-4Vision in neurodegenerative disease histopathology: a comparative analysis with convolutional neural network model
  publication-title: Neuropathol Appl Neurobiol
– volume: 183
  year: 2022
  ident: bib40
  article-title: IDF diabetes atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045
  publication-title: Diabetes Res Clin Pract
– volume: 178
  year: 2024
  ident: bib17
  article-title: A survey on advancements in image–text multimodal models: from general techniques to biomedical implementations
  publication-title: Comput Biol Med
– volume: 154
  start-page: 549
  year: 2012
  end-page: 559.e2
  ident: bib23
  article-title: Nonmydriatic ultrawide field retinal imaging compared with dilated standard 7-Field 35-mm photography and retinal specialist examination for evaluation of diabetic retinopathy
  publication-title: Am J Ophthalmol
– volume: 4
  year: 2024
  ident: bib9
  article-title: A user-friendly approach for the diagnosis of diabetic retinopathy using ChatGPT and automated machine learning
  publication-title: Ophthalmol Sci
– volume: 46
  start-page: 1728
  year: 2023
  ident: 10.1016/j.xops.2025.100911_bib30
  article-title: Artificial intelligence and diabetic retinopathy: AI framework, prospective studies, head-to-head validation, and cost-effectiveness
  publication-title: Diabetes Care
  doi: 10.2337/dci23-0032
– volume: 16
  start-page: 415
  year: 2020
  ident: 10.1016/j.xops.2025.100911_bib36
  article-title: Five-year cost-effectiveness modeling of primary care-based, nonmydriatic automated retinal image analysis screening among low-income patients with diabetes
  publication-title: J Diabetes Sci Technol
  doi: 10.1177/1932296820967011
– volume: 60
  start-page: 9
  year: 2018
  ident: 10.1016/j.xops.2025.100911_bib4
  article-title: Automated screening for diabetic retinopathy – a systematic review
  publication-title: Ophthalmic Res
  doi: 10.1159/000486284
– volume: 25
  year: 2023
  ident: 10.1016/j.xops.2025.100911_bib10
  article-title: Prompt engineering as an important emerging skill for medical professionals: tutorial
  publication-title: J Med Internet Res
  doi: 10.2196/50638
– volume: 263
  start-page: 214
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib7
  article-title: Diagnostic accuracy of artificial intelligence-based automated diabetic retinopathy screening in real-world settings: a systematic review and meta-analysis
  publication-title: Am J Ophthalmol
  doi: 10.1016/j.ajo.2024.02.012
– volume: 124
  start-page: 343
  year: 2017
  ident: 10.1016/j.xops.2025.100911_bib38
  article-title: Automated diabetic retinopathy image assessment software: diagnostic accuracy and cost-effectiveness compared with human graders
  publication-title: Ophthalmology
  doi: 10.1016/j.ophtha.2016.11.014
– volume: 13
  start-page: E140
  year: 2016
  ident: 10.1016/j.xops.2025.100911_bib5
  article-title: Divergent perceptions of barriers to diabetic retinopathy screening among patients and care providers, Los Angeles, California, 2014–2015
  publication-title: Prev Chronic Dis
  doi: 10.5888/pcd13.160193
– volume: 14
  start-page: 454
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib22
  article-title: Effects of prompt engineering on large language model performance in response to questions on common ophthalmic conditions
  publication-title: Taiwan J Ophthalmol
  doi: 10.4103/tjo.TJO-D-23-00193
– volume: 4
  start-page: 290
  year: 2013
  ident: 10.1016/j.xops.2025.100911_bib26
  article-title: Classification of diabetic retinopathy and diabetic macular edema
  publication-title: World J Diabetes
  doi: 10.4239/wjd.v4.i6.290
– volume: 50
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib32
  article-title: Evaluating the efficacy of few-shot learning for GPT-4Vision in neurodegenerative disease histopathology: a comparative analysis with convolutional neural network model
  publication-title: Neuropathol Appl Neurobiol
  doi: 10.1111/nan.12997
– volume: 125
  start-page: 1264
  year: 2018
  ident: 10.1016/j.xops.2025.100911_bib31
  article-title: Grader variability and the importance of reference standards for evaluating machine learning models for diabetic retinopathy
  publication-title: Ophthalmology
  doi: 10.1016/j.ophtha.2018.01.034
– volume: 5
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib11
  article-title: Large language models in ophthalmology: a review of publications from top ophthalmology journals
  publication-title: Ophthalmol Sci
– volume: 17
  start-page: 195
  year: 2021
  ident: 10.1016/j.xops.2025.100911_bib2
  article-title: Current understanding of the molecular and cellular pathology of diabetic retinopathy
  publication-title: Nat Rev Endocrinol
  doi: 10.1038/s41574-020-00451-4
– volume: 10
  start-page: 79
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib13
  article-title: Applications of ChatGPT in the diagnosis, management, education, and research of retinal diseases: a scoping review
  publication-title: Int J Retina Vitr
  doi: 10.1186/s40942-024-00595-9
– volume: 38
  start-page: 2491
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib16
  article-title: Artificial intelligence chatbot interpretation of ophthalmic multimodal imaging cases
  publication-title: Eye
  doi: 10.1038/s41433-024-03074-5
– volume: 15
  year: 2023
  ident: 10.1016/j.xops.2025.100911_bib34
  article-title: Is artificial intelligence the cost-saving lens to diabetic retinopathy screening in Low- and middle-income countries?
  publication-title: Cureus
– volume: 5
  year: 2025
  ident: 10.1016/j.xops.2025.100911_bib14
  article-title: Glaucoma detection and feature identification via GPT-4V fundus image analysis
  publication-title: Ophthalmol Sci
  doi: 10.1016/j.xops.2024.100667
– volume: 120
  start-page: 2587
  year: 2013
  ident: 10.1016/j.xops.2025.100911_bib24
  article-title: Peripheral lesions identified by mydriatic ultrawide field imaging: distribution and potential impact on diabetic retinopathy severity
  publication-title: Ophthalmology
  doi: 10.1016/j.ophtha.2013.05.004
– volume: 108
  start-page: 1443
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib41
  article-title: Comparing generative and retrieval-based chatbots in answering patient questions regarding age-related macular degeneration and diabetic retinopathy
  publication-title: Br J Ophthalmol
  doi: 10.1136/bjo-2023-324533
– volume: 7
  start-page: 4
  year: 2016
  ident: 10.1016/j.xops.2025.100911_bib33
  article-title: Automated detection of diabetic retinopathy in three European populations - university of surrey
  publication-title: J Clin Exp Ophthalmol
  doi: 10.4172/2155-9570.1000582
– volume: 125
  start-page: 1608
  year: 2018
  ident: 10.1016/j.xops.2025.100911_bib3
  article-title: Guidelines on diabetic eye care: the international council of ophthalmology recommendations for screening, follow-up, referral, and treatment based on resource settings
  publication-title: Ophthalmology
  doi: 10.1016/j.ophtha.2018.04.007
– volume: 183
  year: 2022
  ident: 10.1016/j.xops.2025.100911_bib40
  article-title: IDF diabetes atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045
  publication-title: Diabetes Res Clin Pract
  doi: 10.1016/j.diabres.2021.109119
– volume: 16
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib44
  article-title: Evaluating chatgpt's diagnostic accuracy in detecting fundus images
  publication-title: Cureus
– volume: 2
  start-page: e240
  year: 2020
  ident: 10.1016/j.xops.2025.100911_bib37
  article-title: Artificial intelligence for teleophthalmology-based diabetic retinopathy screening in a national programme: an economic analysis modelling study
  publication-title: Lancet Digit Health
  doi: 10.1016/S2589-7500(20)30060-1
– volume: 154
  start-page: 549
  year: 2012
  ident: 10.1016/j.xops.2025.100911_bib23
  article-title: Nonmydriatic ultrawide field retinal imaging compared with dilated standard 7-Field 35-mm photography and retinal specialist examination for evaluation of diabetic retinopathy
  publication-title: Am J Ophthalmol
  doi: 10.1016/j.ajo.2012.03.019
– volume: 127
  start-page: P66
  year: 2020
  ident: 10.1016/j.xops.2025.100911_bib25
  article-title: Diabetic retinopathy preferred practice pattern
  publication-title: Ophthalmology
  doi: 10.1016/j.ophtha.2019.09.025
– volume: 108
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib15
  article-title: Foundation models in ophthalmology
  publication-title: Br J Ophthalmol
  doi: 10.1136/bjo-2024-325459
– volume: 10
  start-page: 307
  year: 2021
  ident: 10.1016/j.xops.2025.100911_bib35
  article-title: Economic evaluations of artificial intelligence in ophthalmology
  publication-title: Asia Pac J Ophthalmol
  doi: 10.1097/APO.0000000000000403
– volume: 9
  year: 2023
  ident: 10.1016/j.xops.2025.100911_bib8
  article-title: How does ChatGPT perform on the united States Medical Licensing Examination (USMLE)? The implications of large Language models for medical education and knowledge assessment
  publication-title: JMIR Med Educ
  doi: 10.2196/45312
– volume: 2
  year: 2025
  ident: 10.1016/j.xops.2025.100911_bib21
  article-title: Evaluation of ChatGPT-4 in detecting referable diabetic retinopathy using single fundus images
  publication-title: AJO Int
  doi: 10.1016/j.ajoint.2025.100111
– volume: 7
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib1
  article-title: Diabetes mellitus, the fastest growing global public health concern: early detection should be focused
  publication-title: Health Sci Rep
  doi: 10.1002/hsr2.2004
– volume: 4
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib9
  article-title: A user-friendly approach for the diagnosis of diabetic retinopathy using ChatGPT and automated machine learning
  publication-title: Ophthalmol Sci
  doi: 10.1016/j.xops.2024.100495
– volume: 108
  start-page: 1384
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib20
  article-title: Unveiling the clinical incapabilities: a benchmarking study of GPT-4V(ision) for ophthalmic multimodal image analysis
  publication-title: Br J Ophthalmol
  doi: 10.1136/bjo-2023-325054
– volume: 12
  year: 2025
  ident: 10.1016/j.xops.2025.100911_bib28
  article-title: Artificial intelligence versus manual screening for the detection of diabetic retinopathy: a comparative systematic review and meta-analysis
  publication-title: Front Med
  doi: 10.3389/fmed.2025.1519768
– volume: 11
  start-page: 48
  year: 2025
  ident: 10.1016/j.xops.2025.100911_bib27
  article-title: The efficacy of artificial intelligence in diabetic retinopathy screening: a systematic review and meta-analysis
  publication-title: Int J Retina Vitr
  doi: 10.1186/s40942-025-00670-9
– volume: 67
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib29
  article-title: Population impact and cost-effectiveness of artificial intelligence-based diabetic retinopathy screening in people living with diabetes in Australia: a cost effectiveness analysis
  publication-title: EClinicalMedicine
  doi: 10.1016/j.eclinm.2023.102387
– volume: 9
  start-page: 71
  year: 2023
  ident: 10.1016/j.xops.2025.100911_bib42
  article-title: Application and accuracy of artificial intelligence-derived large language models in patients with age related macular degeneration
  publication-title: Int J Retina Vitr
  doi: 10.1186/s40942-023-00511-7
– volume: 17
  start-page: 4021
  year: 2023
  ident: 10.1016/j.xops.2025.100911_bib43
  article-title: The utility of ChatGPT in diabetic retinopathy risk assessment: a comparative study with clinical diagnosis
  publication-title: Clin Ophthalmol Auckl NZ
  doi: 10.2147/OPTH.S435052
– volume: 5
  year: 2025
  ident: 10.1016/j.xops.2025.100911_bib12
  article-title: Analysis of ChatGPT responses to ophthalmic cases: can ChatGPT think like an ophthalmologist?
  publication-title: Ophthalmol Sci
  doi: 10.1016/j.xops.2024.100600
– volume: 14
  year: 2019
  ident: 10.1016/j.xops.2025.100911_bib6
  article-title: Systematic review on barriers and enablers for access to diabetic retinopathy screening services in different income settings
  publication-title: PLoS One
  doi: 10.1371/journal.pone.0198979
– volume: 178
  year: 2024
  ident: 10.1016/j.xops.2025.100911_bib17
  article-title: A survey on advancements in image–text multimodal models: from general techniques to biomedical implementations
  publication-title: Comput Biol Med
  doi: 10.1016/j.compbiomed.2024.108709
SSID ssj0002513241
Score 2.3137317
Snippet To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new image...
ObjectiveTo evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new...
Objective: To evaluate the diagnostic accuracy of 4 multimodal large language models (MLLMs) in detecting and grading diabetic retinopathy (DR) using their new...
SourceID doaj
proquest
pubmed
crossref
elsevier
SourceType Open Website
Aggregation Database
Index Database
Publisher
StartPage 100911
SubjectTerms Artificial intelligence
Diabetic retinopathy
Image analysis
Multimodal large language model
Ophthalmology
Ultra-widefield fundus photography
Title Can Multimodal Large Language Models Diagnose Diabetic Retinopathy from Fundus Photos? A Quantitative Evaluation
URI https://www.clinicalkey.com/#!/content/1-s2.0-S266691452500209X
https://www.clinicalkey.es/playcontent/1-s2.0-S266691452500209X
https://dx.doi.org/10.1016/j.xops.2025.100911
https://www.ncbi.nlm.nih.gov/pubmed/41030829
https://www.proquest.com/docview/3256110090
https://doaj.org/article/5a07eef880fa455b8e26860c17f4ac3f
Volume 6
WOSCitedRecordID wos001576792300002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVAON
  databaseName: DOAJ Directory of Open Access Journals
  customDbUrl:
  eissn: 2666-9145
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0002513241
  issn: 2666-9145
  databaseCode: DOA
  dateStart: 20210101
  isFulltext: true
  titleUrlDefault: https://www.doaj.org/
  providerName: Directory of Open Access Journals
– providerCode: PRVHPJ
  databaseName: ROAD: Directory of Open Access Scholarly Resources
  customDbUrl:
  eissn: 2666-9145
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0002513241
  issn: 2666-9145
  databaseCode: M~E
  dateStart: 20210101
  isFulltext: true
  titleUrlDefault: https://road.issn.org
  providerName: ISSN International Centre
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV1Lb9QwELagQohLxZsUqIzEDUXEsR3HJ1TKrji0VXlqb5bj2OoiSFbNLoILv50ZO1m1EtALl8iy7NiaGdufPS9CntfWoRi4vFSNzQWWbO1kLqQupfVwY4ga089H6uSkXiz06YVUX2gTlsIDJ8K9lLZQ3gcQs2CFlE3ty6quCsdUENbxgLtvofSFyxTuwXBqA1Jgo5dMMuj60a8wPncp0TJAM3bpJIoB-y8dSH8DnPHgmd8muyNipAdppnfINd_dJTePR534PbI6tB2NfrTf-hYaHqFtN3zTOyTFZGdfB_ommdR5mkxglo6-R2fnHjMS_6ToZELnG0zjQU_P-nU_vKIH9N3GdtEHDXZEOtuGBb9PPs1nHw_f5mMehdzJUq5zZ7nlvNXcaqa9DKJqQxNg5frCtYw7HSqhC6eroL1QWLJBBGslVLiKNfwB2en6zj8itGlKpV1QuvUAvRyzzvm6dU3ptQcu2Iy8mGhqVilchpnsyL4Y5IBBDpjEgYy8RrJvW2Ko61gBAmBGATBXCUBG-MQ0M3mTwv4HP1r-c2j1p15-GJfwYJgZSlOYD4BgKs2i-hegtV5kRG57jigloY8rR3w2SZSBJYx6Gdv5fjMYDrATI_fpIiMPk6htSSJYDCik9_4HqR6TWzCh8QXpCdlZn2_8U3LDfV8vh_N9cl0t6v24juB7_Gv2G-QWI34
linkProvider Directory of Open Access Journals
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Can+Multimodal+Large+Language+Models+Diagnose+Diabetic+Retinopathy+from+Fundus+Photos%3F+A+Quantitative+Evaluation&rft.jtitle=Ophthalmology+science+%28Online%29&rft.au=Most%2C+Jesse+A&rft.au=Walker%2C+Evan+H&rft.au=Mehta%2C+Nehal+N&rft.au=Nagel%2C+Ines+D&rft.date=2026-01-01&rft.eissn=2666-9145&rft.volume=6&rft.issue=1&rft.spage=100911&rft_id=info:doi/10.1016%2Fj.xops.2025.100911&rft_id=info%3Apmid%2F41030829&rft.externalDocID=41030829
thumbnail_m http://cvtisr.summon.serialssolutions.com/2.0.0/image/custom?url=https%3A%2F%2Fcdn.clinicalkey.com%2Fck-thumbnails%2F26669145%2Fcov200h.gif