Skip to content

Latest commit

 

History

History
81 lines (54 loc) · 3.94 KB

CT-ORG.md

File metadata and controls

81 lines (54 loc) · 3.94 KB

CT-ORG

Dataset Information

CT-ORG (CT volume with multiple organ segmentations) is a CT dataset containing multiple organ segmentations. It consists of 140 CT volumes, among which 131 volumes are paired with PET-CT datasets. The dataset extends the list of organs from the LITS challenge, including 9 additional organs, resulting in a total of 6 categories and 21 organs. The dataset is divided into training, validation, and testing subsets based on LITS, with an emphasis on organs commonly segmented in clinical practice, aiming for a comprehensive organ coverage for CT body scans.

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Anatomical Area Number of Categories Data Volume File Format
3D CT Segmentation Lung, Bone, Liver, Kidney, Bladder, Brain Whole Body 6 119 for training, 21 for test .nii.gz

Resolution Details

Resolution Level spacing (mm) size
Original Image (0.56, 0.56, 0.70) (512, 512, 74)
Medium Resolution (0.78, 0.78, 1.0) (512, 512, 458)
High Resolution (1.37, 1.37, 5.0) (512, 512, 987)

The CT-ORG dataset contains a total of 63,503 images.

Label Information Statistics

Organ Liver Bladder Lungs Kidneys Bone Brain
Case Count 140 114 140 140 140 9
Coverage 100% 80% 100% 100% 100% 6.43%
Min Volume (cm³) 1220 12 875 90 348 340
Median Volume (cm³) 1480 144 2269 317 2499 1432
Max Volume (cm³) 3144 950 8822 494 9485 1516

Visualization

ITK-SNAP Visualization. Red: Liver, Green: Bladder, Dark Blue: Lungs, Yellow: Kidneys, Light Blue: Bone.

File Structure

The official file structure is as follows: images are stored in the form of "volume-x.nii.gz", where x is a one to three-digit case number, ranging from 0 to 139. The first 21 images, numbered 0-20, make up the test set. The remaining images make up the training set. Segmentation data is stored in the form of "labels-x.nii.gz", where x corresponds to the number of the associated image file.

Authors and Institutions

Blaine Rister (Stanford University, Department of Electrical Engineering, USA)

Darvin Yi (Stanford University, Department of Biomedical Data Science, USA)

Kaushik Shivakumar (Stanford University, Department of Biomedical Data Science, USA)

Daniel L. Rubin (Stanford University, Department of Biomedical Data Science & Department of Radiology, USA)

Tomomi Nobashi (Stanford University, Department of Radiology, USA)

Source Information

Official Website: https://github.com/bbrister/ctOrganSegmentation

Download Link: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=61080890

Article Address: https://www.nature.com/articles/s41597-020-00715-8

Publication Date: 2020

Citation

@article{rister2020ct,
  title={CT-ORG, a new dataset for multiple organ segmentation in computed tomography},
  author={Rister, Blaine and Yi, Darvin and Shivakumar, Kaushik and Nobashi, Tomomi and Rubin, Daniel L},
  journal={Scientific Data},
  volume={7},
  number={1},
  pages={381},
  year={2020},
  publisher={Nature Publishing Group UK London}
}

Original introduction article is here.