Is artificial intelligence ready to teach epilepsy through images?

Background: Artificial intelligence (AI) has shown significant potential in various areas of medicine, including neurology. With the increasing availability of data and advances in machine learning algorithms, the question arises: is AI ready to contribute to the teaching and understanding of epilepsy? Epilepsy is a complex neurological condition characterized by recurrent episodes of seizures, the understanding of which is essential for healthcare professionals, patients, and their families. This study investigates the feasibility of using AI to generate illustrative images of epileptic seizures from descriptions provided in command prompts. Objective: To explore the ability of artificial intelligence, specifically language models like GPT-3.5, in conjunction with AI-assisted design tools to generate illustrative images of epileptic seizures from textual descriptions and base images. We aim to assess the quality and usefulness of these generated images in facilitating the understanding of epilepsy by healthcare professionals, patients, and families. Methods: We used the GPT-3.5 language model to generate command prompts based on descriptions of different types of epileptic seizures, including physical, behavioral, and temporal characteristics. Image Generation: We tested AI image generators (Copilot, Getimg.ai, bing.com, shakker, aigenerator) apps to translate the command prompts into illustrative images of epileptic seizures. Results: The generated images were evaluated by a neurophysiologist, PhD MD titular member of the Sociedade Brasileira de Neurofisiologia Clínica (SBNC) to determine images accuracy, clarity, and utility in an educational context. Results Preliminary results indicate that the combination of GPT-3.5 and AI image generators were able to generate illustrative images of epileptic seizures visually informative and similar to reality. Nevertheless, some images were bizarre, inaccurate and deceptive. Besides, in all cases, the research team modified and adjusted GPT-3.5 prompt, while testing image generation, until images were minimally useful. Conclusion: This study demonstrates the potential of artificial intelligence to assist in the teaching and understanding of epilepsy. The generation of illustrative images from textual descriptions offers an innovative and accessible approach to the study of epilepsy. However, further research is needed to validate the accuracy and effectiveness of these images in clinical and educational contexts. The ongoing advancement of AI promises to further improve our ability to teach and manage complex neurological conditions such as epilepsy.