This study explored the ability of the new ChatGPT-4 chatbot to interpret ophthalmic images and answer related questions.

Key Findings:

• ChatGPT-4 achieved an accuracy of ۷۰% in answering multiple-choice questions based on multimodal imaging of ophthalmic cases.
• Performance was better for retinal images (77%) compared to other areas like neuro-ophthalmology (58%).
• The chatbot performed better on non-image-based questions (82%) compared to those requiring image interpretation (65%).

Implications:

• ChatGPT-4 shows potential as an aid for interpreting ophthalmic images, particularly in retinal diseases.
• The model still requires improvement, especially for neuro-ophthalmology and image-heavy questions.
• Further research is needed to ensure reliable integration of such chatbots into clinical practice.

Important Considerations:

• The study used a public dataset which might not reflect real-world complexity.
• Ethical considerations regarding reliance on AI for diagnosis need to be addressed.
• Regulatory frameworks may be necessary for safe clinical implementation.

Overall, this study highlights the potential of AI for ophthalmic image analysis, but also emphasizes the need for ongoing development and responsible use.