ChatGPT in Ophthalmology: Past, Present, and Future

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Large language models (LLMs) are rapidly evolving.
• Benchmarking LLMs in ophthalmology is needed to assess clinical utility.
• Comparing newer LLMs to earlier versions and human experts is crucial.

Purpose of the Study:

• To evaluate the diagnostic accuracy of ChatGPT-4o and ChatGPT-5 for ophthalmic conditions.
• To compare the performance of these newer AI models against ChatGPT-3.5, residents, and specialists.

Main Methods:

• Retrospective cohort study at an academic tertiary medical center.
• Reviewed 63 ophthalmology patient cases from June 2022 to January 2023.
• Created two diagnostic scenarios per patient: history alone (Hx) and history with examination (Hx and Ex).
• Assessed diagnostic accuracy of ChatGPT-4o and ChatGPT-5 for each scenario.

Main Results:

• Using history alone, ChatGPT-5 (73%) and ChatGPT-4o (70%) outperformed ChatGPT-3.5 (54%) and neared resident (75%) and attending (71%) accuracy.
• With examination findings, ChatGPT-5 (94%) and ChatGPT-4o (89%) significantly surpassed ChatGPT-3.5 (68%).
• The latest AI models matched or exceeded human diagnostic performance (residents 94%, attendings 87%).

Conclusions:

• ChatGPT-4o and ChatGPT-5 demonstrate significantly improved diagnostic capabilities in ophthalmology compared to ChatGPT-3.5.
• These advanced AI models achieve diagnostic accuracy comparable to or exceeding that of human clinicians in ophthalmic case diagnosis.