A pilot study of generative AI video for patient communication in radiology and nuclear medicine

Abstract Purpose Effective communication for radiation imaging procedures is critical, especially for patients with literacy or language barriers. Generative Artificial Intelligence (GenAI) provides a new approach for creating personalised, multilingual patient education materials. This pilot study evaluates the effectiveness of GenAI, specifically using HeyGen, in creating personalised patient information videos in the Thai language. Methods We created an avatar of a medical physicist using HeyGen. Two English health information scripts on nuclear medicine and radiology were translated into Thai using HeyGen’s translation tool, and videos were created with the avatar delivering the content in Thai. Thirteen native Thai-speaking medical physicists and postgraduate students evaluated the videos using a 5-point Likert scale, focusing on translation accuracy, naturalness of delivery, and usefulness as a patient education tool. The Bilingual Evaluation Understudy scoring system was used to assess translation quality objectively. Results Both videos received high median scores for translation accuracy (median of 4.0). The BLEU scores were 0.56 and 0.66, indicating good translation quality. Participants reported minor issues with formal language and unnatural phrasing but found the videos understandable and valuable overall. Feedback suggested improving the naturalness of the avatar’s delivery to increase relatability. Conclusions Our pilot study shows that GenAI can effectively create and translate personalised patient information videos into Thai, helping to bridge communication gaps in radiation-related procedures. While minor issues remain, the findings indicate that tools like HeyGen could significantly improve patient communication, particularly for those who face language barriers.