[Notice] Anyone using this dataset, please cite the following challenge overview paper on arXiv.
Luo, Gongning, et al. "Efficient automatic segmentation for multi-level pulmonary arteries: The PARSE challenge." arXiv preprint arXiv:2304.03708 (2023).
2023/02/01 For future researchers, the training and validation sets are open and available. Remember to send the signed document to PARSE2022@hotmail.com for participation. Only after the agreement is received will we the dataset download link be emailed.
Attention, Please! Data Access Rules: The participants should click the Join button, fill out the online registration form and send the signed document to PARSE2022@hotmail.com on the Participation Rules page. After that, we will send you the data download link by email.
Our dataset contains 200 3D volumes with refined pulmonary artery label, these Contrast Enhanced CT Pulmonary Angiography (CTPA) data are obtained from dual-source 64-slice CT scanner in Harbin Medical University, Harbin, China. 10 experts with more than 5-year clinical experience participated in the labeling work. The annotation is performed on the basis of region growing algorithm using MIMICS software.
The image sizes are between 512*512*228and 512*512*376. Pixel sizes of these images are between 0.50mm/pixel
and 0.95mm/pixel, and their slice thicknesses are 1mm/pixel. The images will be stored in .nii.gz files. Voxel-level
segmentation annotations are: 0 - Background, 1 - Pulmonary artery
The proportion of training, validation and test cases is shown as follows:
- Training cases: 100 (The relatively large number of data were used for training a robust model).
- Opened validation cases: 30 (The relatively small number of data were used for validation of algorithm from
different participants to verify the evaluation code by validation dataset and ensure the fairness of the challenge.
At the same time, the relatively small number of data can avoid the disclosure of test set data distribution). - Closed test cases: 70 (The relatively large number of data were used for a fair final leaderboard).