Large language models for patient education prior to interventional radiology procedures: a comparative study
DeepSeek-V3 and ChatGPT-4o excelled on TAPE, BEST, and CT-HDR brachytherapy questions, indicating potential to enhance patient education in interventional radiology, where complex but minimally invasive procedures often are explained in brief consultations. However, OpenBioLLM-8b and BioMistral-7b exhibited more frequent inaccuracies, suggesting that LLMs cannot replace comprehensive clinical consultations yet. Patient feedback and clinical workflow implementation should validate these findings.
Authors
- Tobias Penzkofer (ORCID: https://orcid.org/0000-0001-9591-8575)
- Dirk Schnapauff
- Andrea Dell’Orco (ORCID: https://orcid.org/0000-0002-3964-8360)
- R Schmidt (ORCID: https://orcid.org/0000-0003-0012-2799)
- Willie Magnus Lüdemann (ORCID: https://orcid.org/0000-0002-4255-1009)
- Jawed Nawabi (ORCID: https://orcid.org/0000-0002-1137-0643)
- Anna-Maria Haack
- Semil Eminovic (ORCID: https://orcid.org/0009-0001-1871-2976)
- Bogdan Levita
Institutions
- Max Delbrück Center (DE)
- Humboldt-Universität zu Berlin (DE)
- Berlin Institute of Health at Charité - Universitätsmedizin Berlin (DE)
Publication Details
- Journal
- CVIR Endovascular
- Published
- 2025-10-13
- DOI
- https://doi.org/10.1186/s42155-025-00609-z
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00