The DOI matches its Crossref record.
Comparative Evaluation of Diagnostic and Management Capabilities of Infiniti AI and ChatGPT-4o in Corneal Diseases
Open accessAbstract
BACKGROUND: Artificial intelligence (AI), particularly large language models (LLMs), is rapidly transforming medical education and clinical decision support. Ophthalmology, a specialty heavily reliant on pattern recognition, presents a promising domain for LLM integration. While general-purpose models like ChatGPT-4o have demonstrated strong performance in ophthalmic tasks, domain-specific systems such as Infiniti AI, built with a retrieval-augmented generation (RAG) framework, claim advantages by grounding responses in peer-reviewed ophthalmic literature. This study compares ChatGPT-4o (OpenAI, San Francisco, CA, USA) and Infiniti AI (Sinjab Academy, UAE) in corneal disease case scenarios. MATERIALS AND METHODS: Twenty corneal cases were selected from the University of Iowa EyeRounds database, covering infectious, inflammatory, degenerative, developmental, and systemic associations. ChatGPT-4o, Infiniti AI, and a fellowship-trained cornea specialist independently evaluated each case. Diagnostic and management responses were scored against American Academy of Ophthalmology preferred practice pattern guidelines using a four-point scale (0-3). Statistical comparisons were performed using paired t-tests and Wilcoxon signed-rank tests. RESULTS: ChatGPT-4o significantly outperformed Infiniti AI across all categories. Diagnostic accuracy was higher for ChatGPT-4o (2.37 ± 0.81) than Infiniti AI (1.13 ± 0.71, p < 0.001, Cohen's d = 1.35). Management scores were also superior (2.65 ± 0.65 vs 1.98 ± 0.65, p < 0.001, d = 1.37). Overall, ChatGPT-4o achieved a mean total score of 5.00 ± 1.22 compared with 3.10 ± 1.10 for Infiniti AI (p < 0.001, d = 1.75). CONCLUSIONS: ChatGPT-4o demonstrated greater diagnostic and management accuracy than Infiniti AI in corneal disease scenarios, highlighting the current strength of general-purpose LLMs over specialized retrieval-based systems. Nonetheless, both models remain prone to hallucinations and should serve as adjuncts to, rather than replacements for, expert judgment. Further refinement of ophthalmology-specific models is warranted to improve safety and clinical utility.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In emergency triage and workflow automation, RAG and KG-RAG systems exhibited enhanced performance relative to regular keyword-based or machine-learning triage models.
“In emergency triage and workflow automation, the RAG and KG-RAG systems exhibited enhanced performance relative to regular keyword based or machine-learning triage models [32].”
From Multi-Agent Retrieval Augmented Generation for Clinical Decision Support: A Systematic Review and Integrative Conceptual Framework (2026) · cites this paper as [32]
The Introduction identifies Infiniti AI as a RAG-based ophthalmology model, but the study compares it with ChatGPT-4o on corneal disease cases. The Conclusions report that ChatGPT-4o outperformed Infiniti AI; the paper does not address KG-RAG, emergency triage or workflow automation, or comparisons with keyword-based or machine-learning triage models (searched for those terms). (LLM, conf=0.78)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
No references could be resolved yet.
Research works printed but not linked: 11
- Zhang K, Meng X, Yan X, et al.. Revolutionizing health care: the transformative impact of large language models in medicine. J Med Internet Res. 2025, 27:59069. 10.2196/59069 · DOI
- Fatima A, Shafique MA, Alam K, Fadlalla Ahmed TK, Mustafa MS. ChatGPT in medicine: a cross-disciplinary systematic review of ChatGPT's (artificial intelligence) role in research, clinical practice, education, and patient interaction. Medicine (Baltimore). 2024, 103:39250. 10.1097/MD.0000000000039250 · DOI
- He HJ, Zhao FF, Liang JJ, et al.. Evaluation and comparison of large language models' responses to questions related optic neuritis. Front Med (Lausanne). 2025, 12:1516442. 10.3389/fmed.2025.1516442 · DOI
- Wu JH, Nishida T, Liu TY. Accuracy of large language models in answering ophthalmology board-style questions: A meta-analysis. Asia Pac J Ophthalmol (Phila). 2024, 13:100106. 10.1016/j.apjo.2024.100106 · DOI · Paper page
- Milad D, Antaki F, Milad J, et al.. Assessing the medical reasoning skills of GPT-4 in complex ophthalmology cases. Br J Ophthalmol. 2024, 108:1398-405. 10.1136/bjo-2023-325053 · DOI
- Mikhail D, Milad D, Antaki F, et al.. Multimodal performance of GPT-4 in complex ophthalmology cases. J Pers Med. 2025, 15:4. 10.3390/jpm15040160 · DOI
- Savastano MC, Rizzo C, Fossataro C, et al.. Artificial intelligence in ophthalmology: progress, challenges, and ethical implications. Prog Retin Eye Res. 2025, 107:101374. 10.1016/j.preteyeres.2025.101374 · DOI
- Khake AM, Gokhale S, Dindore P, Khake S, Desai M. Evaluating the use of ChatGPT 3.5 and Bard as self-assessment tools for short answer questions in undergraduate ophthalmology. Cureus. 2025, 17:86288. 10.7759/cureus.86288 · DOI
- Pan X, Bai Y, Zou K, Zhou Y, Zhou J, Fu H. Eh-benchmark: ophthalmic hallucination benchmark and agent-driven top-down traceable reasoning workflow. arXiv. 2025,
- Gilbert S, Kather JN, Hogan A. Augmented non-hallucinating large language models as medical information curators. NPJ Digit Med. 2024, 7:100. 10.1038/s41746-024-01081-0 · DOI
- Waters F, Fernyhough C. Hallucinations: a systematic review of points of similarity and difference across diagnostic classes. Schizophr Bull. 2017, 43:32-43. 10.1093/schbul/sbw132 · DOI
Cited by
Sorted by citations
Mugambiwa, T., & Ndlovu, B. (2026) · Journal of Applied Informatics and Computing · doi:10.30871/jaic.v10i1.11900 · Cited by 4
“Overall, RAG-supported patient education tools have been shown to provide better communication quality and usability compared to traditional static or rule-based educational systems[29][30][31].”
Metadata
- Authors
- Mohammad, A., Bulbanat, A., & Aljassar, F.
- Affiliations
- Farwaniya Hospital (KW)
- Kuwait Institute for Medical Specialization (KW)
- Ibn Sina Hospital (KW)
- Corresponding author
- Mohammad, Abdulaziz
- Venue
- Cureus
- Published
- 2025-10-22
- Volume
- 17
- Issue
- 10
- Pages
- e95163-e95163
- ISSN
- 2168-8184 (ISSN-L)
- PMC
- PMC12543008
- OpenAlex
- W4415435998
- Cited by
- 0
- References
- 0
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- Locations
- Cureus (publishedVersion, cc-by)
- Europe PMC (PubMed Central) (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Topics
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Health Sciences › Medicine › Ophthalmology › Ophthalmology and Visual Health Research
- Health Sciences › Medicine › Ophthalmology › Ocular Infections and Treatments
- Keywords
- corneal disease, large language models, GPT-4o, retrieval-augmented generation, diagnostic accuracy, hallucinations, clinical decision support, ophthalmology
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
