Multi-Class Segmentation of Aortic Branches and Zones in Computed Tomography Angiography
Welcome to the **AortaSeg24** challenge, where innovation meets collaboration in the pursuit of medical imaging excellence.
About the Challenge
This challenge focuses on proposing deep learning models for the segmentation of the Aorta, its branches, and zones. We provide an example script utilizing SwinUNETR to guide you in developing your training script and training a module. Additionally, sample validation code is available to assess your proposed model's performance. Finally, a container submission script is offered to facilitate the submission process to the Grand Challenge. Feel free to modify this script to suit your specific needs.
Dataset Details
Input Image Specifications
- Type: 3D CTA Volume
- Axial Dimensions:
- Minimum: $389\times389$ pixels
- Maximum: $516\times516$ pixels
- Average: $450\times450$ pixels
- Voxel Resolution: Isotropic, uniformly set to $(1mm \times 1mm \times 1mm)$.
- Axial Slice Count:
- Range: 578 to 801 slices
- Average: 695 slices
Output Segmentation Specifications
The output is a 3D segmentation model mirroring the input image characteristics, with the addition of 23 segmentation classes detailed below.
Segmentation Classes
Accessing the Dataset
To access the dataset, please participate in the AortaSeg24 Challenge hosted on the Grand Challenge platform. Begin by visiting the challenge page and proceed to complete the Data Agreement Form. Upon submission and approval, you will gain full access to the dataset.
Citation
If you find our code or data useful in your research, please cite our papers:
-
@article{imran2024cis, title={CIS-UNet: Multi-Class Segmentation of the Aorta in Computed Tomography Angiography via Context-Aware Shifted Window Self-Attention}, author={Imran, Muhammad and Krebs, Jonathan R and Gopu, Veera Rajasekhar Reddy and Fazzone, Brian and Sivaraman, Vishal Balaji and Kumar, Amarjeet and Viscardi, Chelsea and Heithaus, Robert Evans and Shickel, Benjamin and Zhou, Yuyin and Shao, Wei}, journal={arXiv preprint arXiv:2401.13049}, year={2024} }
-
@article{krebs2024volumetric, title={Volumetric Analysis of Acute Uncomplicated Type B Aortic Dissection Using an Automated Deep Learning Aortic Zone Segmentation Model}, author={Krebs, Jonathan R and Imran, Muhammad and Fazzone, Brian and Viscardi, Chelsea and Berwick, Benjamin and Stinson, Griffin and Heithaus, Evans and Upchurch Jr, Gilbert R and Shao, Wei and Cooper, Michol A}, journal={Journal of Vascular Surgery}, year={2024}, publisher={Elsevier} }