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.  


NOTE:
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


Announcements

  • 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!


Overview

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


Sponsorship

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)