DMRI¶

DMRI provides diffusion MRI simulators and tools for model selection and parameter inference. Use it to apply a pretrained model, train a model on simulated data, or evaluate a trained run.

Research use only. DMRI has not been clinically validated and must not be used for clinical decisions. Pretrained checkpoints are acquisition-specific. Confirm that a checkpoint's training acquisition is compatible with your data before interpreting its output; matching filenames is not enough.

Install¶

DMRI requires Python 3.11 or newer. From a clone of the repository:

uv venv -p 3.11
source .venv/bin/activate
uv pip install -e .

See Installation and quickstart for GPU options and a complete prediction example.

Predict¶

Put data.nii.gz, nodif_brain_mask.nii.gz, bvals, and bvecs in one folder, then run:

dmri predict FOLDER --model msb3s_2_4_6_128 --quality balanced --non-interactive

Results and an HTML viewer are written below FOLDER/dmri_output/. See the prediction guide for checkpoint selection, output maps, and performance options.

Other workflows¶

• Train a model
• Evaluate a trained model
• Build and use simulators
• Run the examples
• Look up command options
• Use the Python API
• Check configuration, outputs, and troubleshooting