Scholarlink

<p>Medical illustrations are crucial for education, patient communication, and research, traditionally created by skilled human artists. The emergence of Generative Artificial Intelligence (GAI) models like Midjourney v6.0, DALL-E 3, Gemini Ultra 1.0, Stable Diffusion 2.0, Adobe Firefly, and Microsoft Designer Image Creator presents a transformative shift in this field. These AI models offer significant advantages in terms of speed, cost-effectiveness, and the potential to address diversity gaps in existing illustrations. For instance, studies show GAI’s capacity for rapid image generation and aesthetic quality in craniofacial anatomy. However, current research highlights substantial limitations, primarily concerning anatomical accuracy. Evaluations across craniofacial, foot, and ophthalmic anatomy (specifically corneal transplant procedures) reveal consistent flaws in depicting precise details such as foramina, suture lines, muscle origins, neurovascular structures, and complex bone arrangements. AI-generated images often fall short in anatomical realism, procedural step accuracy, and can even introduce fictitious anatomy when compared to human-created illustrations. These inaccuracies stem from insufficient training data and the models’ incomplete understanding of complex biological structures. Ethical considerations like potential biases and copyright issues also pose challenges. While GAI demonstrates immense promise as a supplemental tool, it currently necessitates rigorous human expert validation for medical reliability. Future developments must focus on enhancing training data quality and fostering interdisciplinary collaboration to fully realize GAI’s potential in creating truly accurate and reliable medical illustrations. Level of Evidence: Not gradable.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial intelligence in medical visualization: a new era for anatomical illustration

  • 期刊论文

  • 发表日期：

    2026年3月12日

查看全文
  • Gabriele Monarchi,
  • Margherita Gobbo,
  • Matteo Val,
  • Giuseppe Consorti,
  • Gabriele Gerolamo Miotti,
  • Luca Guarda Nardini

摘要

Medical illustrations are crucial for education, patient communication, and research, traditionally created by skilled human artists. The emergence of Generative Artificial Intelligence (GAI) models like Midjourney v6.0, DALL-E 3, Gemini Ultra 1.0, Stable Diffusion 2.0, Adobe Firefly, and Microsoft Designer Image Creator presents a transformative shift in this field. These AI models offer significant advantages in terms of speed, cost-effectiveness, and the potential to address diversity gaps in existing illustrations. For instance, studies show GAI’s capacity for rapid image generation and aesthetic quality in craniofacial anatomy. However, current research highlights substantial limitations, primarily concerning anatomical accuracy. Evaluations across craniofacial, foot, and ophthalmic anatomy (specifically corneal transplant procedures) reveal consistent flaws in depicting precise details such as foramina, suture lines, muscle origins, neurovascular structures, and complex bone arrangements. AI-generated images often fall short in anatomical realism, procedural step accuracy, and can even introduce fictitious anatomy when compared to human-created illustrations. These inaccuracies stem from insufficient training data and the models’ incomplete understanding of complex biological structures. Ethical considerations like potential biases and copyright issues also pose challenges. While GAI demonstrates immense promise as a supplemental tool, it currently necessitates rigorous human expert validation for medical reliability. Future developments must focus on enhancing training data quality and fostering interdisciplinary collaboration to fully realize GAI’s potential in creating truly accurate and reliable medical illustrations. Level of Evidence: Not gradable.