The SegTHOR (Segmentation of THoracic Organs at Risk) dataset is a CT dataset specifically for the segmentation of thoracic organs at risk (OARs), which are the organs surrounding the tumor that need to be protected from radiation during radiotherapy, as part of the ISBI 2019 challenge. In this dataset, the risk organs include the heart, trachea, aorta, and esophagus, each differing in spatial and appearance characteristics. The esophagus is manually annotated from the fourth cervical vertebra to the esophagogastric junction, the heart is annotated according to the recommendations of the radiation oncology group, the trachea is annotated from the lower limit of the larynx to 2 cm below the carina but does not include the bronchial branches, and the aorta is annotated from its origin to below the diaphragmatic pillar. The dataset includes 60 CT images, with 40 for training and 20 for testing.
Dimensions | Modality | Task Type | Anatomical Structures | Anatomical Area | Number of Categories | Data Volume | File Format |
---|---|---|---|---|---|---|---|
3D | CT | Segmentation | Heart, Trachea, Aorta, Esophagus. | Chest | 4 | 40 for training, 20 for test. | .nii.gz |
Total number of 2D slices in the 40 training images: 7420.
Dataset Statistics | spacing (mm) | size |
---|---|---|
min | (0.90, 0.90, 2.0) | (512, 512, 147) |
median | (0.98, 0.98, 2.5) | (512, 512, 177) |
max | (1.37, 1.37, 2.5) | (512, 512, 284) |
Segmentation Class | esophagus | heart | trachea | aorta |
---|---|---|---|---|
Case Count | 40 | 40 | 40 | 40 |
Detection Rate | 100% | 100% | 100% | 100% |
Min Volume (cm³) | 30 | 438 | 22 | 95 |
Median Volume (cm³) | 45 | 847 | 38 | 206 |
Max Volume (cm³) | 170 | 1829 | 73 | 465 |
Paper Visualization.
ITK-SNAP Visualization. Red: Heart, Green: Trachea, Blue: Aorta, Yellow: Esophagus.
The SegTHOR dataset includes two folders: train
and test
. The train
folder contains 40 subfolders (from Patient_01
to Patient_40
), each with a label file GT.nii.gz
and a corresponding .nii.gz
data file. The test
folder directly contains 20 .nii.gz
files from Patient_41.nii.gz
to Patient_60.nii.gz
.
SegTHOR/
|-- train/
| |-- Patient_01/
| | |-- GT.nii.gz
| | |-- Patient_01.nii.gz
| |-- Patient_02/
| | ... (similar structure as Patient_01)
| ...
| |-- Patient_40/
| ... (similar structure as Patient_01)
|
|-- test/
|-- Patient_41.nii.gz
|-- Patient_42.nii.gz
...
|-- Patient_60.nii.gz
Caroline Petitjean (University of Rouen, France)
Zoé Lambert (INSA Rouen Normandie, France)
Su Ruan (University of Rouen, France)
Bernard Dubray (University of Rouen, Henri Becquerel Center, Regional Cancer Center, France)
Official Website: https://competitions.codalab.org/competitions/21145
Download Link: https://competitions.codalab.org/competitions/21145#participate-get_starting_kit
Article Address: https://ieeexplore.ieee.org/document/9286453, https://ceur-ws.org/Vol-2349/
Publication Date: January, 2019.
@INPROCEEDINGS{9286453,
author={Lambert, Zoé and Petitjean, Caroline and Dubray, Bernard and Kuan, Su},
booktitle={2020 Tenth International Conference on Image Processing Theory, Tools and Applications (IPTA)},
title={SegTHOR: Segmentation of Thoracic Organs at Risk in CT images},
year={2020},
volume={},
number={},
pages={1-6},
doi={10.1109/IPTA50016.2020.9286453}}
Original introduction article is here.