Dear Participants,

We’re happy to inform you that the Validation Phase is now open! 🚀 Thank you for your patience while we resolved some unexpected technical issues. We have also addressed a few raised issues in the training datasets of Task 2 and 3. The updates are explained below:

Validation Phase 🏁

The Validation Phase consists of two types of algorithm submission: Debug and Development.

  • Debug 🐞

    This is a minimal test for your Docker setup. There’s no limit on submissions here, and the submissions do not count toward your validation quota.

  • Development 📈

    You may now submit your models for validation and hyperparameter tuning. Each team is allowed a maximum of 5 submissions.

  Validation set sizes:

  • Task 1: 23 cases
  • Task 2: 17 cases
  • Task 3: 17 cases

Team registration is required to participate in the Validation Phase.

To get started, we recommend adapting your model using the newly updated minimal baseline. Follow the README instructions to package your model as a Docker image and upload it to the Grand Challenge platform. 📦

Note about Validation Phase 📢

During the week of 23-27th of June, the Task 2 Debug and Development submission pages will be briefly closed for maintenance 🛠️. During which you may experience a few errors interacting with the submission pages. We will make an announcement to inform you once the maintenance has been completed.

Training Data Fixes (Task 2 and 3) ✅

We’ve applied the following corrections:

  • Clinical data files (_CD.json). Previously there were missing variables in some of the files due to a JSON tool's inability to handle missing values (NaN). These missing values are now represented as "-1" (as string or integer).

  • Case 2A_141 (Task 2) and 3A_141 (Task 3). The tissue masks appeared to be flipped. We have corrected the two tissue mask files (_HE_mask.json), and the associated feature and patch coordinate files (_HE.pt, _HE.npy):

    • 2A_141_HE_mask.tif, 2A_141_HE.pt and 2A_141_HE.npy
    • 3A_141_HE_mask.tif, 3A_141_HE.pt and 3A_141_HE.npy

To download only these fixed files:

  • Task 2: aws s3 sync --no-sign-request s3://chimera-challenge/v2/task2/
  • Task 3: aws s3 sync --no-sign-request s3://chimera-challenge/v2/task3/

Use sync (not cp) to download only updated files in your current local training data folder. 📥

  If you have any questions, don’t hesitate to reach out to the CHIMERA team. We’re here to help and in the meantime, enjoy training your models! 🤖

— The CHIMERA Challenge Organizers    

Footnote - Validation phase submission pages:

Minimal baseline (for packaging your model as a docker image): https://github.com/nadieh/CHIMERA_minimal_baseline

Task 1
Task 2
Task 3