HoloCT

3D CT foundation model that takes a CT volume (NIfTI, Hounsfield units) and a text prompt for lesion screening, VQA, report generation and language-guided segmentation, returning 3D masks as evidence for its findings; HoloCT-4B and HoloCT-8B checkpoints on Hugging Face (Nov 2026)

LICMITUPD2026-11-07STATUSMaintainedCOMMPermissive
What it is
Vision-language model (Foundation models & agents)
Modality
Radiology: 3D CT
Maintainer
Alibaba DAMO Academy / Xie et al.
License
MIT
Commercial use
Permissive
Access
Free, public source repository (see links below)
Last update
2026-11-07 (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: Repo LICENSE and both Hugging Face model cards (Alibaba-DAMO-Academy/HoloCT-4B and HoloCT-8B, public, not gated) state MIT for code and weights. THIRD_PARTY_NOTICES lists components from Med3DVLM (MIT), MONAI VISTA (Apache-2.0) and Transformers/Qwen3-VL (Apache-2.0). Training data (HoloDATA) draws on 73 public datasets with their own terms plus clinical cohorts. Needs an NVIDIA GPU with bfloat16. Paper: arXiv 2611.06471 (Nov 4, 2026).

Related tools in Foundation models & agents

← Back to directory