Skip to content

Wingspan Chest Organ Segmentation and Cardiothoracic Ratio (CTR) Estimation Dataset

License

Notifications You must be signed in to change notification settings

eveningdong/cardiothoracic_ratio_estimation_dataset

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Cardiothoracic Ratio (CTR) Estimation Dataset

Description

Extended Wingspan Cardiothoracic Ratio (CTR) Estimation Dataset used in the paper Unsupervised Domain Adaptation for Automatic Estimation of Cardiothoracic Ratio and Neural Architecture Search for Adversarial Medical Image Segmentation. The extension includes additional 38 patients and chest organ segmenation masks for all patients.

The dataset contains 259 patients' chest X-ray images, chest organ segmentation masks, and key points for CTR estimation.

Full Dataset

The full dataset can be downloaded vis this link.

Data Structure

Each patient has an independent folder with the folder name as his ID. In each folder, there are original.dcm, left_lung.png, right_lung.png, heart.png, and one key_points.txt.

In key_points.txt, you will see name,x,y. name is the organ name, and x and y are the normalized pixel coordinates for the width and height. (x=0, y=0) stands for the upper left corner and (x=1, y=1) stands for the lower right corner.

Downsampled Dataset

The downsampled dataset with fixed resolution 512 x 512 is provided within the repository. In addition, another popular chest organ segmentation dataset Japanese Society of Radiological Technology(JSRT) is provided here as a comparison.

Data Structure

Two datasets wingspan and jsrt have the same structure. Each dataset has two folders png and mask, where mask has three sub-folders left_lung, right_lung and heart.

Python API

Python functions for CTR estimation is provided in utils.py.

Citation

If you find the work and dataset are useful in your research, please cite:

@inproceedings{dong2018unsupervised, title={Unsupervised domain adaptation for automatic estimation of cardiothoracic ratio}, author={Dong, Nanqing and Kampffmeyer, Michael and Liang, Xiaodan and Wang, Zeya and Dai, Wei and Xing, Eric}, booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention}, pages={544--552}, year={2018}, organization={Springer} }

@inproceedings{dong2019neural, title={Neural architecture search for adversarial medical image segmentation}, author={Dong, Nanqing and Xu, Min and Liang, Xiaodan and Jiang, Yiliang and Dai, Wei and Xing, Eric}, booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention}, pages={828--836}, year={2019}, organization={Springer} }

About

Wingspan Chest Organ Segmentation and Cardiothoracic Ratio (CTR) Estimation Dataset

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages