AI in Hand and Wrist Radiography: Multimodal Large Language Models for Distal Radius Fracture Detection and Characterization
: Only two models exceeded 90% sensitivity for fracture detection, while intra-articular extension remained at chance level (≤55.6%). Substantial inter-run reliability (κ > 0.60) was observed in only one model. These findings indicate that current MLLMs do not reliably support multidimensional fracture assessment and that single-run evaluations overestimate robustness.
Authors
- Armin Kraus (ORCID: https://orcid.org/0000-0001-6557-2163)
- Gerrit Grieb (ORCID: https://orcid.org/0000-0002-7302-210X)
- David Breidung (ORCID: https://orcid.org/0000-0002-2966-2863)
- Henrik Stelling
- Martin Lautenbach
- Ibrahim Güler
Institutions
- Witten/Herdecke University (DE)
- Friedrich-Alexander-Universität Erlangen-Nürnberg (DE)
- University Hospital of Zurich (CH)
- Krankenhaus Waldfriede (DE)
- Gemeinschaftskrankenhaus Havelhöhe (DE)
- Berlin Center for Epidemiology and Health Research (DE)
- RWTH Aachen University (DE)
- Otto-von-Guericke-Universität Magdeburg (DE)
Publication Details
- Journal
- Diagnostics
- Published
- 2026-04-15
- DOI
- https://doi.org/10.3390/diagnostics16081171
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00