3D DMFNet for Real-time Brain Tumor Segmentation

This repository is the work of "3D Dilated Multi-Fiber Network for Real-time Brain Tumor Segmentation in MRI" based on pytorch implementation. You could click the link to access the paper. The multimodal brain tumor dataset (BraTS 2018) could be acquired from here.

Dilated multi-fiber network

Architecture of 3D DMFNet

Requirements

python 3.6
pytorch 0.4 or 1.0
nibabel
pickle

Implementation

Download the BraTS2018 dataset and change the path:

experiments/PATH.yaml

Data preprocess

Convert the .nii files as .pkl files. Normalization with zero-mean and unit variance .

python preprocess.py

(Optional) Split the training set into k-fold for the cross-validation experiment.

python split.py

Training

Sync bacth normalization is used so that a proper batch size is important to obtain a decent performance. Multiply gpus training with batch_size=8 is recommended. The total training time take less than 10 hours in gtxforce 1080Ti.

python train_all.py --gpu=0,1,2,3 --cfg=DMFNet_GDL_all --batch_size=8

Test

Table. Dice scores for the enhancing tumor, whole tumor and tumor core respevtively

Model	Params (M)	FLOPs (G)	Dice_ET (%)	Dice_WT (%)	Dice_TC (%)
0.75x MFNet	1.81	13.36	79.34	90.22	84.25
MFNet	3.19	20.61	79.91	90.43	84.61
DMFNet	3.88	27.04	80.12	90.62	84.54

Where ET: the enhancing tumor, WT: the whole tumor, TC: the tumor core.

You could download the trained DMFnet (pytorch) from Google drive or Baidu dirve(password for download: t8xu). Put the weight in the dir './ckpt/DMFNet_GDL_all/here'. You could obtain the resutls as paper reported by running the following code:

python test.py --mode=1 --is_out=True --verbose=True --use_TTA=True --postprocess=True --snapshot=True --restore=model_last.pth --cfg=DMFNet_GDL_all --gpu=0

Then make a submission to the online evaluation server.

Citation

If you use our code or model in your work or find it is helpful, please cite the paper:

@inproceedings{chen2019dmfnet,
  title={3D Dilated Multi-Fiber Network for Real-time Brain Tumor Segmentation in MRI},
  author={Chen, Chen and Liu, Xiaopeng and Ding, Meng and Zheng, Junfeng and Li, Jiangyun},
  booktitle={International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI)},
  year={2019}
}

Luoxd1996/BraTS-DMFNet