Generic research assets
Convenience mirrors of third-party models and templates. These are not original models trained by the repository owner and are not official upstream releases.
The original licenses must be respected. Hosting here does not replace, supersede or waive any original license, attribution, citation or redistribution requirement. There is no single blanket license for this collection. Consult LICENSES.md and the accompanying license files before use or redistribution. No endorsement by the original authors is implied. Research use only; this mirror provides no clinical approval, warranty or guarantee of segmentation accuracy.
TumorSynth portable assets
| Path | Contents | Original terms |
|---|---|---|
tumorsynth/whole-tumor-v1/ |
Unmodified Task002 five-fold final checkpoints, checkpoint metadata, plans and postprocessing | UCL TumorSynth Software License Agreement |
templates/sri24/spgr.nii |
Unmodified SRI24 SPGR registration template | CC BY-SA 3.0 |
synthstrip.1.pt |
Existing main SynthStrip v1 checkpoint, retained without replacement | Authors offer MIT or CC BY 4.0; CC BY 4.0 documented here |
SHA256SUMS covers these inference assets; provenance.json records sources and exact file sizes. No patient scans or segmentation outputs are added with this bundle. Checkpoint pickle files must only be loaded from trusted sources. Pin a Hugging Face commit revision for reproducibility.
Original sources and papers
- TumorSynth, Jiaming Wu and Ferran Prados Carrasco: project/code, documentation and original weight links, whole-tumor archive. Wu et al., TumorSynth: Integrated Brain Tumor and Tissue Segmentation in Brain MRI Scans of any Resolution and Contrast, Radiology: Imaging Cancer (2026), listed as accepted in the upstream references. The upstream link currently points to the journal, not an article DOI.
- SRI24: original atlas distribution. Rohlfing et al., The SRI24 Multichannel Atlas of Normal Adult Human Brain Structure, Human Brain Mapping 31:798–819 (2010), doi:10.1002/hbm.20906.
- SynthStrip: original code, weights and licensing. Hoopes et al., SynthStrip: Skull-Stripping for Any Brain Image, NeuroImage 260:119474 (2022), doi:10.1016/j.neuroimage.2022.119474.
- nnU-Net, the TumorSynth inference architecture/framework: code, Isensee et al., nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation, Nature Methods (2021), doi:10.1038/s41592-020-01008-z.
Other existing assets
The existing IXI example and MNI152 templates are unchanged. Their respective original dataset/template terms still apply; the license statements above do not relicense them. Consult IXI and FSL standard templates.