segresnet-mlx

Segment brain tumors, renal structures and whole-body CT anatomy with MONAI SegResNet models in native MLX.

Docs · MedMLX

BraTS bundle output on a Medical Segmentation Decathlon brain MRI: tumor core (red), edema (green) and enhancing tumor (yellow).

Apple Silicon · Metal · Python 3.13 / 3.14 · uv

Quick start

uv venv --python 3.13 --seed
source .venv/bin/activate
python -m pip install "segresnet-mlx==0.2.1" huggingface-hub

The tagged GitHub release is an alternative installation route:

repo="https://github.com/MedMLX/segresnet-mlx.git@v0.2.1"
python -m pip install "segresnet-mlx @ git+$repo" huggingface-hub

Download the public converted BraTS bundle separately, then run:

from huggingface_hub import hf_hub_download
from segresnet_mlx.medmlx import run

bundle = "brats_mri_segmentation"
weights = hf_hub_download("MedMLX/segresnet-brats-mlx", "segresnet-brats.safetensors")

result = run(
    bundle=bundle,
    t1c_path="t1c.nii.gz",
    t1_path="t1.nii.gz",
    t2_path="t2.nii.gz",
    flair_path="flair.nii.gz",
    weights_path=weights,
    output_dir="out",
)
print(result["outputs"]["mask"])

Keep input volumes on the same voxel grid. The package contains code only. Public converted bundles are available separately for BraTS and renal CECT; follow each weight card and its terms. Docs cover local upstream checkpoint conversion and whole-body CT bundles. Qualification differs by bundle; see the technical notes.

License

Apache-2.0. See third-party notices. Model weights retain their publisher terms.

For research and education. Outputs require human review.