Revolution or risk?—Assessing the potential and challenges of GPT-4V in radiologic image interpretation
Question Can Generative Pre-trained Transformer 4 Vision (GPT-4V) interpret radiologic images-with and without clinical context? Findings GPT-4V performed poorly, demonstrating diagnostic accuracy rates of 8% (uncontextualized), 29% (contextualized, most likely diagnosis correct), and 64% (contextualized, correct diagnosis among differential diagnoses). Clinical relevance The utility of commercial multimodal large language models, such as GPT-4V, in radiologic practice is limited. Without clinical context, diagnostic errors and fabricated findings may compromise patient safety and misguide clinical decision-making. These models must be further refined to be beneficial.
Authors
- Can Yüksel (ORCID: https://orcid.org/0000-0002-1719-9875)
- Daniel Truhn (ORCID: https://orcid.org/0000-0002-9605-0728)
- Robert Siepmann (ORCID: https://orcid.org/0000-0002-4608-5180)
- Christiane Katharina Kuhl (ORCID: https://orcid.org/0000-0001-8696-2363)
- Marc Huppertz (ORCID: https://orcid.org/0009-0004-2801-3679)
- Omid Nikoubashman (ORCID: https://orcid.org/0000-0002-2055-4217)
- Sven Nebelung (ORCID: https://orcid.org/0000-0002-5267-9962)
- David Topp
Institutions
- Universitätsklinikum Aachen (DE)
- RWTH Aachen University (DE)
Publication Details
- Journal
- European Radiology
- Published
- 2024-10-18
- DOI
- https://doi.org/10.1007/s00330-024-11115-6
- Citations
- 38
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 4.57
Funders
- European Commission
- Deutsche Forschungsgemeinschaft
- Bundesministerium für Bildung und Forschung
- HORIZON EUROPE Framework Programme
- HORIZON EUROPE European Research Council