SegMamba

Now we have open-sourced the pre-processing, training, inference, and metrics computation codes.

SegMamba: Long-range Sequential Modeling Mamba For 3D Medical Image Segmentation

https://arxiv.org/abs/2401.13560

Our advantage in speed and memory.

Contact

If you have any questions about our project, please feel free to contact us by email at [email protected] or via WeChat at 18340097191.

Environment install

Clone this repository and navigate to the root directory of the project.

git clone https://github.com/ge-xing/SegMamba.git

cd SegMamba

Install causal-conv1d

cd causal-conv1d

python setup.py install

Install mamba

cd mamba

python setup.py install

Install monai

pip install monai

Simple test

python 0_inference.py

Preprocessing, training, testing, inference, and metrics computation

Data downloading

Data is from https://arxiv.org/abs/2305.17033

Download from Baidu Disk https://pan.baidu.com/s/1C0FUHdDtWNaYWLtDDP9TnA?pwd=ty22提取码ty22

Download from OneDrive https://hkustgz-my.sharepoint.com/:f:/g/personal/zxing565_connect_hkust-gz_edu_cn/EqqaINbHRxREuIj0XGicY2EBv8hjwEFKgFOhF_Ub0mvENw?e=yTpE9B

Preprocessing

In my setting, the data directory of BraTS2023 is : "./data/raw_data/BraTS2023/ASNR-MICCAI-BraTS2023-GLI-Challenge-TrainingData/"

First, we need to run the rename process.

python 1_rename_mri_data.py

Then, we need to run the pre-processing code to do resample, normalization, and crop processes.

python 2_preprocessing_mri.py

After pre-processing, the data structure will be in this format:

Training

When the pre-processing process is done, we can train our model.

We mainly use the pre-processde data from last step: data_dir = "./data/fullres/train"

python 3_train.py

The training logs and checkpoints are saved in: logdir = f"./logs/segmamba"

Inference

When we have trained our models, we can inference all the data in testing set.

python 4_predict.py

When this process is done, the prediction cases will be put in this path: save_path = "./prediction_results/segmamba"

Metrics computation

We can obtain the Dice score and HD95 on each segmentation target (WT, TC, ET for BraTS2023 dataset) using this code:

python 5_compute_metrics.py --pred_name="segmamba"

Acknowledgement

Many thanks for these repos for their great contribution!

https://github.com/MIC-DKFZ/nnUNet

https://github.com/Project-MONAI/MONAI

https://github.com/hustvl/Vim

https://github.com/bowang-lab/U-Mamba