RAD-CaseBookLLM-08

RAD-CaseBookLLM-08 is an open-access dataset containing Large Language Model (LLM)-generated educational texts focused on radiological differential diagnosis. The dataset was generated using ChatGPT-4o (OpenAI, web-based version, March 2025) under standardized prompting conditions (new user account, conversation memory disabled, new chat session for each topic). For each entry, a structured prompt was used, varying only the radiological theme under study. Prompts and responses were written exclusively in English, and all outputs were copied verbatim without editing, preserving formatting and model-generated concluding statements. The thematic inputs provided to the LLM correspond to radiological “key imaging findings” topic titles from the casebook Top 3 Differentials in Radiology: A Case Review (O’Brien WT, 2010). No copyrighted text, images, figures, case descriptions, explanations, or other protected material from the original publication were reproduced, copied, or included in this dataset. All educational content contained herein was independently generated by the LLM based solely on the thematic titles. The dataset is organized by radiology subspecialty and provided in both PDF and DOCX formats, distributed as compressed ZIP archives. It was created as part of a multicenter comparative study evaluating the perceived educational usefulness of LLM-generated differential diagnosis teaching material versus a traditional radiology casebook among junior and advanced radiology trainees. The primary purpose of this dataset is to provide a structured LLM-generated equivalent of a radiology differential diagnosis casebook covering diverse thematic imaging findings. It is intended to serve as a research resource for studying the educational characteristics, strengths, limitations, and reproducibility of LLM-generated medical teaching content. This dataset is not designed or validated for direct clinical use or as formally accredited educational material. This dataset is released under the CC0 1.0 Universal license to promote transparency, reproducibility, and further research in radiology education and medical artificial intelligence. A comprehensive methodological description of the dataset will be provided in a dedicated data publication.