Generative Methods for Pediatric Genetics Education.

Area of Science:

• Medical Education
• Artificial Intelligence
• Genetics

Background:

• Generative AI, including tools like ChatGPT and DALL-E, is increasingly used across various fields.
• There is growing interest in applying generative AI to medical education for specialized training.
• This study investigates the efficacy of generative AI in training pediatric residents to identify genetic syndromes.

Approach:

• Generative AI methods were used to create novel images of individuals with Kabuki syndrome (KS) and Noonan syndrome (NS) from existing datasets.
• Generated images were validated for accuracy using an external classifier.
• Pediatric residents (n=208) evaluated 20 images each after one of four educational interventions, comparing real versus AI-generated images.

Key Points:

• Generative AI images performed similarly to real images in diagnostic training but were perceived as slightly less helpful.
• The majority of participants found the images useful for learning, with a preference for real images.
• Real images were considered more beneficial for educational purposes compared to AI-generated ones.

Conclusions:

• Generative AI-created images can function as an effective adjunctive educational resource in medical training.
• This approach may be particularly beneficial for recognizing less common or unfamiliar genetic conditions like KS.
• Further research can explore optimizing generative AI tools for enhanced medical education outcomes.