Skip to content

Latest commit

 

History

History
139 lines (102 loc) · 7.73 KB

AbdomenAtlas_Mini.md

File metadata and controls

139 lines (102 loc) · 7.73 KB
Raw

AbdomenAtlas 1.0 Mini

Dataset Information

AbdomenAtlas 1.0 Mini is a subset of the AbdomenAtlas-8K dataset that is currently publicly available, including 5,195 detailed annotated CT volumetric data. The AbdomenAtlas-8K dataset itself is much larger, containing a total of 8,448 CT cases. AbdomenAtlas Mini covers nine categories (eight different organs, but the left and right kidneys are marked as separate organs in the annotations): spleen, liver, left kidney, right kidney, stomach, gallbladder, pancreas, aorta, and inferior vena cava (IVC). It is also worth noting that the original image data for AbdomenAtlas Mini comes from 16 public datasets depicted in the figure below, most of which have already been detailed in previous articles; a summary table of these datasets can be found at the end of the referenced articles.

Dataset # of organs # of volumes # of centers source countries license # of volumes in AbdomenAtlas-8K
Pancreas-CT [74] 1 82 1 US CC BY 3.0 42
LiTS [7] 1 201 7 DE, NL, CA, FR, IL CC BY-SA 4.0 131
KiTS [29] 1 300 1 US CC BY-NC-SA 4.0 300
AbdomenCT-1K [54] 4 1,112 12 DE, NL, CA, FR, IL, US, CN CC BY-NC-SA 1000
CT-ORG [72] 5 140 8 DE, NL, CA, FR, IL, US CC BY 3.0 140
CHAOS [86] 4 40 1 TR CC BY-SA 4.0 20
MSD CT Tasks [1] 9 947 1 US CC BY-SA 4.0 945
BTCV [45] 12 50 1 US CC BY 4.0 47
AMOS22 [39] 15 500 2 CN CC BY-NC-SA 200
WORD [50] 16 150 1 CN GNU GPL 3.0 120
FLARE’23 13 4,000 30 - CC BY-NC-ND 4.0 2200
AbdomenCT-120rgan [53] 12 50 - - CC BY-NC-SA 50
  • The table comes from the appendix of the original paper. Note that the last column represents the number of contributions each public dataset makes to the AbdomenAtlas Mini dataset.

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Anatomical Area Number of Categories Data Volume File Format
3D CT Segmentation Spleen, liver, left and right kidneys, stomach, gallbladder, pancreas, aorta, inferior vena cava Abdomen 9 5195 .nii.gz

Resolution Details

Statistics based on the 5170 cases of data we actually downloaded.

Dataset Statistics spacing (mm) size
min (0.44, 0.39, 0.36) (29, 79, 49)
median (0.83, 0.81, 2.0) (493, 364, 205)
max (5.0, 3.0, 8.0) (1059, 519, 2572)

Number of two-dimensional slices in the data set: 1,638,917

Label Information Statistics

Category Occurrence Percentage Median Volume (cm³) Maximum Volume (cm³)
aorta 5162 99.84% 76 611
gallbladder 5122 99.07% 20 159
kidney_left 5139 99.40% 172 1609
kidney_right 5135 99.33% 173 1690
liver 5166 99.92% 1531 8555
pancreas 5162 99.84% 77 1327
postcava 5165 99.90% 68 366
spleen 5155 99.70% 216 3088
stomach 5162 99.84% 263 2411

Visualization

Official Visualization.

File Structure

The file structure of the AbdomenAtlas1.0Mini dataset is as follows: it contains several numbered folders, such as BDMAP_00000001, each containing a compressed three-dimensional CT scan image (ct.nii.gz) of a patient. Under each case folder, the segmentations subdirectory includes multiple .nii.gz format files, which detail annotations for 9 major abdominal organs.

AbdomenAtlas1.0Mini
├── BDMAP_00000001
│   ├── ct.nii.gz
│   └── segmentations
│       ├── aorta.nii.gz
│       ├── gall_bladder.nii.gz
│       ├── kidney_left.nii.gz
│       ├── kidney_right.nii.gz
│       ├── liver.nii.gz
│       ├── pancreas.nii.gz
│       ├── postcava.nii.gz
│       ├── spleen.nii.gz
│       └── stomach.nii.gz
├── BDMAP_00000002
├── BDMAP_00000003
├── ...
└── BDMAP_00005195

Authors and Institutions

Chongyu Qu (Johns Hopkins University, USA)

Tiezhang Zhang (Johns Hopkins University, USA)

Hualin Qiao (Rutgers University, USA)

Jie Liu (City University of Hong Kong)

Yucheng Tang (NVIDIA Corporation, USA)

Alan L. Yuille (Johns Hopkins University, USA)

Zongwei Zhou (Johns Hopkins University, USA)

Source Information

Official Website: https://github.com/MrGiovanni/AbdomenAtlas

Download Link: https://huggingface.co/datasets/AbdomenAtlas/AbdomenAtlas_1.0_Mini

Article Address: https://www.cs.jhu.edu/~alanlab/Pubs23/qu2023abdomenatlas.pdf

Publication Date: 2024-02

Citation

@article{qu2023abdomenatlas,
  title={Abdomenatlas-8k: Annotating 8,000 CT volumes for multi-organ segmentation in three weeks},
  author={Qu, Chongyu and Zhang, Tiezheng and Qiao, Hualin and Tang, Yucheng and Yuille, Alan L and Zhou, Zongwei and others},
  journal={Advances in Neural Information Processing Systems},
  volume={36},
  year={2023}
}

@inproceedings{li2024well,
  title={How Well Do Supervised Models Transfer to 3D Image Segmentation?},
  author={Li, Wenxuan and Yuille, Alan and Zhou, Zongwei},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024}
}

Original introduction article is here.