Concerns regarding image orientation and annotation completeness

Hi all,

I’ve been examining the dataset closely and wanted to raise a couple of observations that might be worth discussing:

Image Orientation: Some of the scans appear to be inconsistently oriented.

Annotation Completeness: The annotations for the tumor seem to be limited to sparse 2D slices in several cases, rather than full 3D segmentations. And I didn't even see any associated annotations for a patient.

Will there be an update to the dataset, or should we work with this dataset? Looking forward to hearing others' thoughts, and would appreciate any clarification from the organizers if possible.

Thanks!

Hi Cédric,

Thank you for your observations and for taking the time to examine the dataset so thoroughly.

We're currently working on addressing the issues you mentioned, including image orientation inconsistencies and the completeness of tumor annotations. A new release of the dataset will be published shortly to correct these aspects. This update will also include the STAPLE-based consensus segmentation that will be used for evaluating the binary segmentation task.

We appreciate your patience and will update the challenge page once the revised version is available.

Best regards, The Organizing Team

Dear Organizing Team,

Thank you for your prompt response and for the clarification.

It's great to hear that these issues are being addressed and that a revised dataset release is on the way. We look forward to working with the updated version.

Best regards, Cédric