VerSe: Large Scale Vertebrae Segmentation Challenge
Look well to the spine for the cause of disease - Hippocrates
Table of Contents
What is VerSe?
Spine or vertebral segmentation is a crucial step in all applications regarding automated quantification of spinal morphology and pathology. With the advent of deep learning, for such a task on computed tomography (CT) scans, a big and varied data is a primary sought-after resource. However, a large-scale, public dataset is currently unavailable.
We believe VerSe can help here. VerSe is a large scale, multi-detector, multi-site, CT spine dataset consisting of 374 scans from 355 patients. The challenge was held in two iterations in conjunction with MICCAI 2019 and 2020. The tasks evaluated for include: vertebral labelling and segmentation.
If you use VerSe, we would appreciate references to the following papers.
Sekuboyina A et al., VerSe: A Vertebrae Labelling and Segmentation Benchmark for Multi-detector CT Images, 2021.
In Medical Image Analysis: https://doi.org/10.1016/j.media.2021.102166
Löffler M et al., A Vertebral Segmentation Dataset with Fracture Grading. Radiology: Artificial Intelligence, 2020.
In Radiology AI: https://doi.org/10.1148/ryai.2020190138
Liebl H and Schinz D et al., A Computed Tomography Vertebral Segmentation Dataset with Anatomical Variations and Multi-Vendor Scanner Data, 2021.
The dataset has four files corresponding to one data sample: image, segmentation mask, centroid annotations, a PNG overview of the annotations.
- 01_training - Train data
- 02_validation - (Formerly) PUBLIC test data
- 03_test - (Formerly) HIDDEN test data
Sub-directory-based arrangement for each patient. File names are constructed of entities, a suffix and a file extension following the conventions of the Brain Imaging Data Structure (BIDS; https://bids.neuroimaging.io/)
Example: ------- training/rawdata/sub-verse000 sub-verse000_dir-orient_ct.nii.gz - CT image series training/derivatives/sub-verse000/ sub-verse000_dir-orient_seg-vert_msk.nii.gz - Segmentation mask of the vertebrae sub-verse000_dir-orient_seg-subreg_ctd.json - Centroid coordinates in image space sub-verse000_dir-orient_seg-vert_snp.png - Preview reformations of the annotated CT data.
- Centroid coordinates of the subject based structure (.json file) are given in voxels in the image space. 'label' corresponds to the vertebral label:
- 1-7: cervical spine: C1-C7
- 8-19: thoracic spine: T1-T12
- 20-25: lumbar spine: L1-L6
- 26: sacrum - not labeled in this dataset
- 27: cocygis - not labeled in this dataset
- 28: additional 13th thoracic vertebra, T13
Note: The annotation format of the complete VerSe data is NOT identical to the one used for the MICCAI challenges. The OSF repositories above also point to the MICCAI version of the data and annotations. Nonetheless, we recommend usage of the restructured data and annotations
The data is provided under the CC BY-SA 4.0 License, making it fully open-sourced.
The rest of this repository is under the MIT License.
We provide helper code and guiding notebooks.
- Data reading, standardising, and writing: Data utilities
- Evaluation (as employed in the 2020 challenge): Evaluation utilities
- Notebooks: Data preperation, Evaluation
Other Related Work
VerSe has resulted in numerous other publications. Below are a few selected ones.
Sekuboyina A. et al., Labelling Vertebrae with 2D Reformations of Multidetector CT Images: An Adversarial Approach for Incorporating Prior Knowledge of Spine Anatomy. Radiology: Artificial Intelligence, 2020. (https://doi.org/10.1148/ryai.2020190074)
- This work is supported by the European Research Council (ERC) under the European Union’s ‘Horizon2020’ research & innovation programme (GA637164–iBack–ERC–2014–STG).
- We thank NVIDIA for the support in challenge organisation with compute.
- README derived from PCam.