Data Description

train/MR: The training MR images

train/Mask: The mask of training MR images. Each vertebra or intervertebral disc was assigned a unique label. The label values and the corresponding spinal structures are:

0: Background

1: S

2: L5

3: L4

4: L3

5: L2

6: L1

7: T12

8: T11

9: T10

10: T9

11: L5/S

12: L4/L5

13: L3/L4

14: L2/L3

15: L1/L2

16: T12/L1

17: T11/T12

18: T10/T11

19: T9/T10

where T, L, and S denote thoracic, lumbar, and sacral vertebra respectively.

There are 172 training data in the preliminary competition, including MR images and mask labels, 20 test data in the preliminary competition and 23 test data in the  second round competition. The labels of the preliminary competition testset and the second round competition testset are not published, and the results can be evaluated online on this website.

Spine Label

Note: the data disclosed in this competition belongs to Southern Medical University. The data is only used for participating in this competition and scientific research. It is prohibited to be used for commercial purposes without permission.
Although the competition is over, the data can still be downloaded.
To download the dataset, you must sign the dataset use license agreement and send the agreement to the e-mail: 

A data download link will be sent after review.

Use this dataset,please refer to and cite the article: “Shumao Pang, Chunlan Pang, Lei Zhao, Yangfan Chen, Zhihai Su, Yujia Zhou, Meiyan Huang, Wei Yang, Hai Lu, Qianjin Feng*. SpineParseNet: Spine Parsing for Volumetric MR Image by a Two-Stage Segmentation Framework with Semantic Image Representation [J]. IEEE Transactions on Medical Imaging, 2021, 40(1): 262-273.”