RadFinder
3D CT vision-language model for report-to-scan retrieval, prompt-based disease classification and slice-level localization; RadFinder-2610 checkpoints released Oct 22, 2026
LICMIT (code); CC BY-NC-SA 4.0 (weights)UPD2026-10-22STATUSMaintainedCOMMNon-commercial
- What it is
- Vision-language model (Foundation models & agents)
- Modality
- Radiology: 3D CT (chest, abdomen-pelvis)
- Maintainer
- University of Freiburg (LMB) / Ging et al.
- License
- MIT (code); CC BY-NC-SA 4.0 (weights)
- Commercial use
- Non-commercial
- Access
- Free, public source repository (see links below)
- Last update
- 2026-10-22 (last repository push, UTC; Maintained)
- Last checked
- 2026-11-16
- Limitations
- Listed for research and development reference. Hounsfield Lab has not tested its clinical performance; check the project’s own documentation for intended use and regulatory status.
Links
License review: Code is MIT, except files in src/rate/ vendored from YalaLab/rate, which are ECL-2.0. Hugging Face weights (lmb-freiburg/radfinder-2610-full, -refct, -public, -ctrate) are CC BY-NC-SA 4.0 and also subject to the training datasets' licenses (e.g. CT-RATE is CC BY-NC-SA 4.0). Published at MICCAI 2026.
Commercial use: read the project’s license (and any model-weight terms) yourself before relying on it. This isn’t legal advice.
Related tools in Foundation models & agents
- Pillar-0 (finetune code + HF weights): Open 3D radiology foundation model; finetuning framework, weights on Hugging Face
- Pillar-0 pretraining: Vision-language pretraining code for Pillar-0
- RATE: Uses an LLM (Qwen3 via sglang) to extract structured labels for 366 findings from radiology reports
- Merlin: 3D vision-language model for CT that uses structured EHR and report supervision
- MedGemma: Google's open medical model family; repo has notebooks and code
- CheXagent: Foundation model for chest X-ray interpretation