Brain2Vec
Self-supervised 3D masked-autoencoder foundation model pretrained on 74,425 T1-weighted brain MRIs; used as a fine-tuning backbone or frozen feature extractor (e.g., brain age, disease classification)
- What it is
- Foundation model (Neuroimaging)
- Modality
- Neuro: structural T1-weighted brain MRI
- Maintainer
- UPenn CBICA (Davatzikos lab)
- License
- CBICA non-commercial agreement (code); CC BY-NC-ND 4.0 (weights)
- Commercial use
- Non-commercial
- Access
- Free, public source repository (see links below)
- Last update
- 2026-11-06 (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: The repo LICENSE points to the UPenn CBICA non-commercial software agreement (https://www.med.upenn.edu/cbica/software-agreement-non-commercial.html). Weights on Hugging Face (cbica/Brain2Vec) are gated with manual approval and tagged CC BY-NC-ND 4.0. Inputs must be skull-stripped and linearly registered to the provided template; the repo added a preprocessing pipeline for this on Oct 27, 2026 (NiChart_DLMUSE skull-stripping, then FSL flirt). Paper: arXiv 2610.28411 (Oct 20, 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 Neuroimaging
- HD-BET: Deep-learning brain extraction (skull stripping) for MRI
- BrainChop: In-browser 3D MRI rendering and segmentation
- FreeSurfer: Neuroimaging analysis and visualization suite
- fMRIPrep: Robust preprocessing pipeline for fMRI
- DIPY: Python library for diffusion MRI, tractography and registration
- nilearn: Machine learning for neuroimaging in Python