Generative AI for Patient Communication in Radiology and Nuclear Medicine: A Pilot Study in Thai

Abstract

Purpose: Effective patient communication in procedures involving radiation is crucial, especially for patients facing literacy or language barriers. Generative Artificial Intelligence offers innovative solutions for creating personalised, multilingual patient education materials. This pilot study assesses the effectiveness of GenAI, specifically using HeyGen, in generating personalised patient information videos in the Thai language. Methods: An avatar of a medical physicist was created using HeyGen. Two English health information scripts on nuclear medicine and radiology were translated into Thai with HeyGen's translation tool, and videos were generated featuring the avatar delivering the content in Thai. Native Thai-speaking medical physicists and postgraduate students (n = 13) evaluated the videos using a 5-point Likert scale on criteria such as translation accuracy, naturalness of delivery, and usefulness as a patient education tool. Objective translation quality was assessed using the Bilingual Evaluation Understudy scoring system. Results: Both videos received high median scores for translation accuracy (median = 4.0), with BLEU scores of 0.57 and 0.66, indicating good translation quality. Participants noted minor issues with formal language and unnatural phrasing but generally found the videos understandable and valuable. Feedback suggested improvements in the naturalness of the avatar's delivery to enhance relatability. Conclusions: This pilot study shows that GenAI can effectively create personalised patient information videos and translate them into Thai, helping to bridge communication gaps in procedures involving radiation. While minor issues remain, the findings suggest that tools like HeyGen could significantly support patient communication with further refinement, especially for those facing language barriers.

Ähnliche Arbeiten