Related Experiment Video
Updated: May 5, 2026
07:51
Full-Field Optical Coherence Microscopy for Histology-Like Analysis of Stromal Features in Corneal Grafts
Published on: October 21, 2022
2.2K
Evaluating the Accuracy and Readability of ChatGPT-4o's Responses to Patient-Based Questions about Keratoconus
Ali Safa Balci1, Semih Çakmak2
1Department of Ophthalmology,Sehit Prof. Dr. Ilhan Varank Sancaktepe Training and Research Hospital, Sehit Prof. Dr. Ilhan Varank Sancaktepe Training and Research Hospita University of Health Sciences, Istanbul, Türkiye.
Ophthalmic Epidemiology
|March 28, 2025
Summary
ChatGPT-4o provides accurate answers to keratoconus patient questions, but its complex language may hinder understanding. Further improvements are needed for AI medical content readability.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Informatics
Background:
- Keratoconus patient education is crucial for management.
- Large language models (LLMs) like ChatGPT-4o are increasingly used for health information.
- Evaluating the accuracy and readability of LLM responses in ophthalmology is essential.
Purpose of the Study:
- To assess the accuracy of ChatGPT-4o responses to patient-centered keratoconus questions.
- To evaluate the readability of AI-generated keratoconus information.
- To compare accuracy and readability across different question types.
Main Methods:
- ChatGPT-4o answered 30 patient-centered keratoconus questions.
- Responses were scored for accuracy (1-5) by two ophthalmologists.
- Readability assessed using SMOG, FKGL, and FRE scores.
Main Results:
- Mean accuracy score was 4.48 ± 0.57 (strong interrater reliability).
- Readability scores indicated a high educational level required (SMOG: 15.49, FKGL: 14.95, FRE: 27.41).
- Accuracy did not differ by question type, but treatment questions were most readable.
Conclusions:
- ChatGPT-4o offers highly accurate keratoconus information.
- Current AI-generated medical content may lack accessibility for the general public.
- Enhancing LLM readability for medical information is a key area for future development.
