项目作者: JohnleeHIT

项目描述 :
Multi-step cascaded network for brain tumor segmentations (tensorflow)
高级语言: Python
项目地址: git://github.com/JohnleeHIT/Brats2019.git
创建时间: 2019-08-09T03:04:53Z
项目社区:https://github.com/JohnleeHIT/Brats2019

开源协议:MIT License

下载


Multi-step cascaded network for brain tumor segmentations with tensorflow

This is an implementation of our BraTS2019 paper “Multi-step Cascaded Networks for Brain Tumor segmentation” on Python3, tensorflow, and Keras.




Whole Tumor……………………Tumor Core ………………….Enhancing Tumor

Schematic of the proposed method



Requirements

Python3.5, Tensorflow 1.12 and other common packages which can be seen in requirements.txt

Getting started

  • parameters.ini is the configuration file.
  • main.py is the main function of the program.
  • oprations.py builds the model and is responsible for the training and the testing phase.
  • utils.py provides lots of useful functions which is utilized in the program.
  • models.py provides some combinations of multiple layers and model achitecture.
  • seg_eval.py provides some APIs to calculate the metrics, e.g. DIce, Sensitivity, etc.

Training

The network is trained on the Brain Tumor Segmentation Challenge 2019(Brats2019) training dataset which can be downloaded from Brats2019 web page .

(1) Edit parameters.ini so as to be consistent with your local environment, especially the “phase”, “traindata_dir “ and “testdata_dir “, for example:

  1. phase = train
  2. traindata_dir = /home/xxx/Dataset/BraTS2019_train/train
  3. testdata_dir = /home/xxx/Dataset/mix/test

notice : folder structure of the training or testing data should be like this:

train/test——-HGG/LGG——BraTS19_XXX_X_X—-BraTS19_XXX_X_X_flair.nii.gz

​ —-BraTS19_XXX_X_X_t1.nii.gz

​ —-BraTS19_XXX_X_X_t1ce.nii.gz

​ —-BraTS19_XXX_X_X_t2.nii.gz

(2) Run main.py in the command line or in the python IDE directly.

  1. python main.py

Testing

(1) Edit parameters.ini so as to be consistent with your local environment, especially the “phase”, “traindata_dir “ and “testdata_dir “, for example:

  1. phase = test
  2. traindata_dir = /home/xxx/Dataset/BraTS2019_train/train
  3. testdata_dir = /home/xxx/Dataset/mix/test

(2) Run main.py in the command line or in the python IDE directly.

  1. python main.py

Results

The proposed method was validated on the Brats2019 evaluation platform, the preliminary results on training and validation sets are as follows:

Dice Coefficient WT TC ET
Training 0.915 0.832 0.791
Validation 0.886 0.813 0.771

To better illustrate the results of the proposed method, we made a qualitative analysis of the segmentation results, which can be seen as follows:



Troubleshooting

If you meet any questions when you run this code , please don’t hesitate to raise a new issue in the repository or directly contact us at lxycust@gmail.com.

Citations

Please consider citing this project in your publications if it helps your research. The following is a BibTeX reference. The BibTeX entry requires the url LaTeX package.

@inproceedings{li2019multi, title={Multi-step cascaded networks for brain tumor segmentation}, author={Li, Xiangyu and Luo, Gongning and Wang, Kuanquan}, booktitle={International MICCAI Brainlesion Workshop}, pages={163--173}, year={2019}, organization={Springer} }