AbdomenCT-1K

License

CC BY 4.0
Creative Commons Attribution 4.0 International License

Citation

Paper BibTeX:

@article{ma2021abdomenct,
  title={Abdomenct-1k: Is abdominal organ segmentation a solved problem?},
  author={Ma, Jun and Zhang, Yao and Gu, Song and Zhu, Cheng and Ge, Cheng and Zhang, Yichi and An, Xingle and Wang, Congcong and Wang, Qiyuan and Liu, Xin and others},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  volume={44},
  number={10},
  pages={6695--6714},
  year={2021},
  publisher={IEEE}
}

Dataset description

AbdomenCT-1K is a large, diverse abdominal CT organ segmentation dataset with over 1,000 scans from 12 medical centers, covering multiple phases, vendors, and diseases. It serves as a benchmark to reveal and address the limited generalization of state-of-the-art methods, providing tasks for fully, semi-, weakly supervised, and continual learning research.

Challenge homepage: https://abdomenct-1k-fully-supervised-learning.grand-challenge.org/

Number of CT volumes: 1062

Contrast: Contrast-enhanced (multi-phase: plain, arterial, portal)

CT body coverage: Abdomen

Does the dataset include any ground truth annotations?: Yes

Original GT annotation targets: Liver, spleen, kidney, pancreas

Number of annotated CT volumes: 1000

Annotator: Initial model + manual refinement

Acquisition centers: 12 medical centers

Pathology/Disease: Lesions in one or more labeled organs, including benign/malignant liver lesions and cancers of the pancreas, colon, and liver

Original dataset download link:

Part 1: https://zenodo.org/records/5903099

Part 2: https://zenodo.org/records/5903846

Part 3: https://zenodo.org/records/5903769

Original dataset format: nifti