Automated cerebral microbleed segmentation for T2*/SWI MRI.
git clone https://github.com/zxKyouma/MedNet-CMB.git
cd MedNet-CMB
conda env create -f environment.yml
conda activate MedNet-CMB
pip install -e .Download the trained model (~1.2 GB) from the latest release.
tar -xzf MedNet-CMB_weights.tar.gz # creates MedNet-CMB_weights/nnUNet_results/You also need SynthStrip's model file for preprocessing:
curl -LO https://github.com/zxKyouma/MedNet-CMB/releases/download/v1.0.0/synthstrip.1.ptIt's redistributed here for convenience from FreeSurfer, which is where it comes from and whose licence it stays under. If you cite this tool, cite SynthStrip too (Hoopes et al., NeuroImage 2022).
Name each scan <subject_id>_0000.nii.gz, put them in one folder, and:
WEIGHTS_DIR=/path/to/MedNet-CMB_weights \
INPUT_DIR=/path/to/raw_images \
OUTPUT_DIR=/path/to/output \
SYNTHSTRIP_MODEL=/path/to/synthstrip.1.pt \
bash run_MedNet-CMB.shThis command does preprocessing and segmentation in one go.
| Cohort | Cluster Dice | Precision | Sensitivity | False Positives / scan |
|---|---|---|---|---|
| Internal (n=264) | 0.82 | 0.88 | 0.77 | 0.58 |
| AIBL | 0.89 | 0.87 | 0.94 | - |
| ASPREE | 0.58 | 0.68 | 0.55 | - |
| VALDO | 0.70 | 0.76 | 0.69 | - |
| External DOU (n=20) | 0.76 | 0.83 | 0.79 | 0.95 |