Native medical imaging on Apple Silicon, built with MLX. Segmentation, detection, synthesis and CT reasoning, running locally on your Mac.
Website · Hugging Face · Python API, CLI & MCP
Requires Apple Silicon, Metal, Python 3.13 and uv.
uv tool install --python 3.13 "medmlx[all,mcp]"
medmlx modelsUse medmlx describe <model> for its inputs and examples. See the package's
setup guide for current
automatic-weight support and source-only release features.
- Request a model to run on your Mac.
- Ask a question about installation, model selection or usage.
- Report a bug in the affected package.
- Share your work with MedMLX.
| Repository | Purpose |
|---|---|
| mlx-reason-ct | CT reports and questions |
| flexible-unet-mlx | Endoscopic tool segmentation |
| swin-unetr-mlx | Abdominal CT segmentation |
| segresnet-mlx | CT and brain MRI segmentation |
| fdg-pet-ct-mlx | Synthetic FDG PET |
| radimagenet-mlx | Radiology image features |
| unest-mlx | Renal CT segmentation |
| biomedparse-mlx | Text-prompted medical segmentation |
| nnunet-mlx | nnU-Net v2 inference |
| totalsegmentator-mlx | Whole-body CT segmentation |
| lung-nodule-mlx | Lung nodule detection |
| synthseg-mlx | Brain segmentation |
| mlx-segment-ct | Prompted CT segmentation |
| flexray-mlx | Radiograph segmentation |
| medsam2-mlx | Prompted masks and propagation |
| lesiondiffusion-mlx | Synthetic CT lesions |
| mlx-generate-ctmr | Synthetic CT and MRI |
| mlx-radiomics | Radiomics and voxel maps |
| med-vlm-mlx | Medical vision-language models |
For research and education. Outputs require human review. Code, model weights and datasets retain their own licenses; see each repository’s notices.
