Comparison of anatomy image generation capability in AI image generation models

Area of Science:

• Medical Illustration
• Artificial Intelligence
• Anatomical Imaging

Background:

• High-performing AI image generation models have emerged, impacting various fields.
• Accurate anatomical depiction is critical in medical contexts, distinguishing from purely creative applications.

Purpose of the Study:

• To evaluate AI models' ability to generate accurate anatomical images.
• To assess AI's understanding of anatomical terminology and accuracy across bones, organs, and muscles.
• To compare initial AI models with their advanced versions released a year later.

Main Methods:

• Assessed AI models (DALL-E 2, Midjourney 5, Stable Diffusion 1.5) for anatomical terminology learning and image accuracy.
• Categorized evaluations across bones, organs, and muscles.
• Conducted a follow-up comparison with updated models (Midjourney 6, DALL-E 3).

Main Results:

• Initial AI models demonstrated varying capabilities in generating accurate anatomical images.
• Advanced AI versions showed improvements but still exhibited limitations in precise anatomical representation.
• AI models struggled with nuanced anatomical details compared to human expertise.

Conclusions:

• AI image generation models cannot fully substitute the skills of professional medical illustrators.
• AI can function as a complementary tool to enhance anatomical education and medical communication.
• The integration of AI is poised to reshape traditional medical illustration practices.