For the current study, a multidisciplinary, multicenter team of radiation protection experts, including radiologists, medical physicists, and radiographers from Switzerland (Centre Hospitalier Universitaire Vaudois, CHUV), the United States (Duke University Medical Center), and Lebanon (Centre Hospitalier Universitaire Libanais Geitaoui) retrieved five of the twelve queries tested on experts in the previous study(4) among those with the lowest variance. A questionnaire was then developed, starting with demographic questions, one of the five selected radiation protection questions, and responses generated by both ChatGPT and a human expert.

Since November 2024, the questionnaire has been randomly distributed to patients in the radiology waiting room at CHUV. An expert approaches the patients to obtain oral informed consent, after which they are invited to complete the questionnaire. Participants can either scan a QR code or use a paper form to complete the questionnaire.

Blinded to the source of the response, participants are asked to evaluate both ChatGPT-generated and human-generated answers based on understandability, trust, reassurance, satisfaction (using the QUEST framework)(8), and their perception of whether the response was AI-generated. All ratings are recorded on a 7-point Likert scale.

Statistical analysis includes a Wilcoxon test with Bonferroni correction to compare scores across these five parameters. Recruitment is ongoing in Lebanon and the United States. Depending on the final data distribution, a regression analysis will be conducted to assess the influence of demographic factors such as gender, age, education, prior exposure to imaging tests, and chatbot use.