label discrepancies

Hello, I understand that the challenge focuses on detecting cases of pancreatic cancer (PDAC) and that other pancreatic lesions should be ignored. However, I noticed that in the training and development sets, three cases contain lesions in the ground-truth segmentation mask but are labeled as "non-PDAC" in the clinical information spreadsheet. Is this discrepancy intentional or an error? Will the test set include such cases with non-PDAC lesions? Thank you in advance! * Cases 100598_00001, 100667_00001, and 101632_00001

Hi Noelie,

Thank you for pointing out the label discrepancies. We reviewed those cases and they are all negative (non-PDAC) cases. The voxel-level labels have been corrected and uploaded to the GitHub repository: https://github.com/DIAGNijmegen/panorama_labels.

Don't hesitate to reach out if you encounter any other issues.

Hi,

I want to ask about case 101290_00001_0000, is the label correct or no?

Best

Hello,

The label for case 101290 has been manually revised and updated in the GitHub repository.

Thank you for bringing up this issue. Let me know if you have any further questions.

Kind regards,

Natalia Alves

Hi,

Case: 100019_00001 has different image shape (1024 x 1024 x 224), and label shape (512 x 512 x 151).

Best

Hello,

We have checked this on our side, and case 100019_00001 has shape (512x512x151), matching the label shape. Could you double-check the issue at your end?

Best,

Natalia Alves

Hi,

I apologize, I wrote the wrong ID.

Here are IDs with mismatched pixel spacings between images and labels: 1. 100028_00001 2. 100936_00001

Best

Hello,

We have looked into this but couldn't reproduce the issue on our side. Please make sure you are using the most updated labels that can be found in this repository: https://github.com/DIAGNijmegen/panorama_labels.

Regards,

Natália