Performance of large language models for CAD-RADS 2.0 classification derived from cardiac CT reports
LLMs enhanced with in-context learning are capable of autonomously generating CAD-RADS 2.0 scores for cardiac CT reports with excellent accuracy, potentially enhancing both the efficiency and consistency of cardiac CT reporting. Open-source models not only deliver competitive accuracy but also present the benefit of local hosting, mitigating concerns around data security.
Authors
- Ákos Varga‐Szemes (ORCID: https://orcid.org/0000-0002-2781-7462)
- Dmitrij Kravchenko (ORCID: https://orcid.org/0000-0003-0331-2045)
- Fabian Bamberg (ORCID: https://orcid.org/0000-0002-7460-3942)
- Martin Soschynski (ORCID: https://orcid.org/0000-0001-6314-1829)
- M. Vecsey-Nagy (ORCID: https://orcid.org/0009-0005-6294-0051)
- Maximilian Frederik Russe (ORCID: https://orcid.org/0000-0003-3187-2429)
- Muhammad Taha Hagar (ORCID: https://orcid.org/0000-0001-9826-2372)
- Philipp Arnold (ORCID: https://orcid.org/0000-0003-0686-7361)
- Tilman Emrich (ORCID: https://orcid.org/0000-0003-4156-7727)
- Alexander Rau (ORCID: https://orcid.org/0000-0001-5881-6043)
- Elmar Kotter (ORCID: https://orcid.org/0000-0001-9022-6000)
- Ayaat Ashi
Institutions
- Semmelweis University (HU)
- Medical University of South Carolina (US)
- University of Freiburg (DE)
- Johannes Gutenberg University Mainz (DE)
- University Hospital Bonn (DE)
- University Medical Center Freiburg (DE)
- University Medical Center of the Johannes Gutenberg University Mainz (DE)
- Cardiff University (GB)
- Cardiff Metropolitan University (GB)
Publication Details
- Journal
- Journal of cardiovascular computed tomography
- Published
- 2025-04-09
- DOI
- https://doi.org/10.1016/j.jcct.2025.03.007
- Citations
- 7
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 4.58
Funders
- Deutsche Forschungsgemeinschaft
- Albert-Ludwigs-Universität Freiburg