/atlasnet

This is the implementation of the pytorch version of AtlasNet.

Primary LanguagePython

image

AtlasNet

This repository is the official implementation of Deep Atlas Network for Efficient 3D Left Ventricle Segmentation on Echocardiography which is presented at Medical Image Analysis 2020.

[paper][code]

Set up the the environment

conda env create -n atlasnet -f atlasnet.yaml
conda activate atlasnet

Data

In our experiments, we used the following dataset:

We use the CETUS dataset in our code, and the image size was changed to 128x128x128.

In the folder ./CETUS, we use some examples to show the format of the data.

Config

To change the default config, modify this file: config

Training

To train our model, run this command:

python train.py

Test

To test the trained our model, run:

python test.py

Citation

@inproceedings{DONG2020101638,
  title={Deep Atlas Network for Efficient 3D Left Ventricle Segmentation on Echocardiography},
  author={Suyu Dong and Gongning Luo and Clara Tam and Wei Wang and Kuanquan Wang and Shaodong Cao and Bo Chen and Henggui Zhang and Shuo Li},
  journal = {Medical Image Analysis},
  volume = {61},
  pages = {101638},
  year = {2020},
  issn = {1361-8415},
}