VerSe`19: Large Scale Vertebrae Segmentation Challenge

The Large Scale Vertebrae Segmentation Challenge (VerSe2019) has been accepted to be held in conjunction with MICCAI 2019, Shenzen, China.  

Requests to join the challenge will no longer be approved*. Registration will be re-opened after the challenge (13.Oct.2019)

* as on 05 Aug, 2019


  • 07 Aug `19: Test data for Test Phase 1 (public) released! (Download from the 'Data' page)

  • 30 July `19: VerSe has been selected by NVIDIA to sponsor a GPU for the challenge winner! (More details soon...)
  • 24 June `19: General announcement: You can access the performance on individual scans on the 'Results' page. Just click on your 'mean position'  to view the detailed JSON performance file.  
  • 17 June `19: Data split has bee revised. New split: Training: 80 scans. Testing (public): 40 scans. Testing (hidden): 40 scans.
  • 17 June `19: Training data (Phase 3) released! (Download from the 'Data' page)
  • 30 June `19: The discussion forum for you to talk about anything and everything related to the challenge is now open. Please check the 'Forum' page.
  • 11 June `19: Training data (Phase 2) released! (Download from the 'Data' page)
  • 1 June `19: Postponement of Training data (Phase 2): The release of the second phase of data has been tentatively postponed to 10 June, 2019.  
  • 17 May `19: Training data (Phase 1) released! (Download from the 'Data' page). More details regarding the two class of algorithms: fully-automated and interactive, will follow soon.
  • 16 May `19: The data host is facing an unexpected IT-related issue due to the file size. We are currently fixing this. Once fixed, the registration requests will be approved and the Phase 1 for training will be made available. Thank you for your patience!


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.

Key highlights of this challenge include:
  • Public release of a spine dataset with an unprecedented 150 (+50 hidden) 120 (+40 hidden) CT scans with voxel-level vertebral annotations.
  • Two sub-tasks, evaluated and ranked separately: (1) Vertebral localisation and identification. (2) Vertebral segmentation.
  • Submissions can be fully-automated and semi-automated/interactive algorithms.
  • Best-performing in each sub-tasks and of all varieties, along with off-beat approaches will be collected, analysed, and presented in a journal article co-authored with the contributors.
  • Contributors retain the intellectual rights to the submitted codes. The organisers will use it solely towards analysis and manuscripts related to the challenge.

Important Dates:

  • Training data release:
    Phase 1 (16 May, 2019)
    Phase 2 (11 June, 2019)
    Phase 3 (17 July, 2019)
  • Test data release: 01 07 Aug, 2019
  • Submission opens (Test Phase 1 ): 01 07 Aug, 2019
  • Submission deadline (Test Phase 1): 15  21 Aug, 2019
  • Docker Submission Opens (Test Phase 2): 01  07 Aug, 2019
  • Docker Submission Closes (Test Phase 2): 30 06 Sep, 2019


This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No  637164 — iBack — ERC-2014-STG)