296 posters, 7 videos, 13 audios, 14 topics, 10 sessions, 1,019 authors, 260 institutions
ePostersLive by SciGen Technologies S.A. All rights reserved.
RCOphth Annual Congress 2026
18 - 21 May, 2026 | Manchester Central, Manchester
P214
Can Artificial Intelligence Platforms Diagnose Eye Disease � A comparison between Chat GPT-4o mini and super prompt agent Alan
Poster Presenter
Authors
Part of Topic
Miscellaneous
Can Artificial Intelligence Platforms Diagnose Eye Disease?
A comparison between Chat GPT-4o mini and super-prompt agent Alan
Introduction
Eye disease encompasses a range of conditions that can affect individuals'quality of life (1). Artificial intelligence(AI) platforms could provide solutions to address the eye care needs of populations that are inadequately provisioned (2).This study's objective was to compare two AI platforms (Alan vs Chat GPT-4o mini) on their outputs in response to clinical eye disease cases.
Aims
Compare each platform based on:
-
Differential diagnoses
-
Referral urgency
-
Language used
When prompted with eye disease cases based on the World Health Organizations(WHO) Primary Eye Care Manual (PEC).
Method
Clinical eye disease cases made from the WHO PEC manual domains:Loss of vision, Trauma, Paediatrics, Lumps and lashes and Red eye. 5 cases per domain equalled 25 cases in total. Each case had 3 presenting complaints labelled A, B and C. This allowed for 3 case variations to be made AB, BC and AC. Both platforms were prompted with each case variation three times. Analysis was performed based on the platforms suggested diagnosis, management, and referral outcomes in comparison to gold standard answers from a consultant ophthalmologist. Additionally, readability of the transcripts was determined by performing a Flesch-Kincaid(FK) analysis; an online readability calculator.
Results
-
Significant differences between the diagnostic accuracies of the platforms was indicated through statistical testing.
-
Referral suggestions from the correctly diagnosed cases were correct in 39/55 for Alan and 30/35 for ChatGPT-4o mini, with no statistical significance found between referral suggestions.
-
Paired t-tests showed highly statistically significant linguistic differences between platforms, most notably between the lower average reading ease and word count per transcript from Alan in comparison to ChatGPT-4o mini.
Conclusion
This study aims to build on existing research into the abilities of AI as a diagnostic tool and has shown more favourable outcomes to tailored platforms. Integration of specialised technologies into primary eye care and teleophthalmology could reduce the global burden of eye disease.
References
1. World Health Organization (2023) Blindness and visual impairment. Available at: https://www.who.int/news-room/fact-sheets/detail/blindness-and-visual-impairment
2. Ting DSJ, Foo VH, Yang LWY, et al Artificial intelligence for anterior segment diseases: Emerging applications in ophthalmology British Journal of Ophthalmology 2021;105:158-168.