SiNGR

SiNGR: Brain Tumor Segmentation via Signed Normalized Geodesic Transform Regression

Accepted as a paper conference at MICCAI 2024

Arxiv: Link

Dependencies

We provide file env.yaml for dependencies. If generating the Geodesic files is needed, use pip install -r requirements_geo.txt.

Alternatively, it is possible to use fast implementation of Geodesic: pip install FastGeodis.

Installation

conda env create -f env.yaml
conda activate singr
pip install -e .

Split data

cd mlpipeline/utils
python split_brats.py
python split_lggflair.py

Applying SiNG transform

cd mlpipeline/utils
python geodesic_transform.py --label_name "fast_sgc_margin" \
      --dataset ${DATASET} \
      --root_dir ${ROOT_DIr} \
      --gt_dir ${GT_DIR} \
      --output_dir ${OUTPUT_DIR}

Training

python -m mlpipeline.train.run experiment=${EXP_NAME} \        
        model.params.cfg.arch=${ARCH_NAME}

where

• <EXP_NAME>: experiment setting can be brats_uncertainty_sem_seg, brats_sem_seg, lgg_uncertainty_sem_seg, or lgg_sem_seg.
• <ARCH_NAME>: architecture name can be EoFormer, Unet3D, UnetPP3D, SegResNet, UNETR, SwinUNETR, TransBTS, or NestedFormer.

Evaluation

python -m mlpipeline.train.evaluate \
    --config=${EXP_NAME} \
    --output_dir=/path/to/inference_results/${EXP_NAME} \
    --log_dir=/path/to/eval/${EXP_NAME} \
    --visual_dir=/path/to/visuals \
    --metadata_path=/path/to/test_split.pkl \
    --dataset_name=${DATASET} \
    --seeds=${SEEDS} \
    --folds=0

where

• DATASET: is either brats or lgg.