Dall-E in hand surgery: Exploring the utility of ChatGPT image generation

Area of Science:

• Medical Education
• Artificial Intelligence in Surgery
• Hand Surgery

Background:

• Artificial intelligence (AI) is increasingly impacting medical fields, including plastic and hand surgery.
• Large language models (LLMs) like ChatGPT and text-to-image generators (e.g., DALL-E, GPT-4o) are becoming more prevalent.
• This study investigates the utility of these AI tools for patient and medical education in hand surgery.

Purpose of the Study:

• To evaluate the capabilities and limitations of AI chatbots (ChatGPT-4.5, ChatGPT-3.5) and text-to-image generators (GPT-4o, DALL-E 3, DALL-E 2) in hand surgery education.
• To assess the accuracy, understandability, and actionability of AI-generated content.
• To compare the performance of different AI models in generating educational materials.

Main Methods:

• Utilized Google Trends data to identify common hand surgery search terms across five categories.
• Queried ChatGPT-4.5 and ChatGPT-3.5 with identified terms and evaluated responses using DISCERN, PEMAT, and Flesch-Kincaid Grade Level.
• Assessed visual representations generated by GPT-4o, DALL-E 3, and DALL-E 2 based on selected ChatGPT responses.

Main Results:

• ChatGPT-4.5 demonstrated superior performance over ChatGPT-3.5 across all metrics, achieving a DISCERN score of 3.80, PEMAT understandability of 91.67%, and accuracy of 4.47.
• AI-generated text had an average Flesch-Kincaid Grade Level of 9.26, indicating a high school reading level.
• GPT-4o generated more accurate visuals compared to DALL-E 3 and DALL-E 2.

Conclusions:

• ChatGPT-4.5 and GPT-4o show potential for enhancing hand surgery education through combined text and image generation.
• Current AI tools possess strengths and weaknesses that require further development for reliable clinical application.
• Additional refinement of AI technologies is necessary prior to their widespread integration into surgical education and practice.