Public Training and Development Dataset: Updates and Fixes

The PANORAMA: Public Training and Development Dataset will consist of 2238 multi-center, multi-vendor cases. Learn more about all the considerations that we've made to curate and release the all-new largest public training dataset for PDAC detection in CECT: panorama.grand-challenge.org/datasets-imaging-labels/. Please monitor this thread for all updates and fixes regarding this dataset.

Imaging data will be released via: zenodo.org/uploads/10959620. Annotations will be released and are maintained via: github.com/DIAGNijmegen/panorama_labels.

HI,

clicking on this link, https://github.com/DIAGNijmegen/panorama_labels, I see error 404, page not found. Is it my problem or does this page still not exist?

Thank you

Hello, the image page appears as follows: Permission required.

Dear Giuseppe,

Thank you for your message. You should try to log-in to GitHub before trying to access the link. After you logged-in, you will see the page and you should (for now) have to ask for permission to access the page; we will try to make to page easier accessible in the future.

Hopefully this will work out for you!

Kind regards,

Megan Schuurmans

Dear Hongkun,

Thank you for your message. You are correct; for now, you have to ask permission to access the page. We are trying to make this easier accessible in the future.

Hopefully this answers your question!

Kind regards,

Megan Schuurmans

Dear Megan, Thank you for the reply. I tried logging into github with my credentials. After logging in and clicking on the link I get "404, page not found". Do I have to apply to access the page? Maybe it's my problem. Thank you for your support. Best regards

Hi, thank you for updating the dataset! Can I know the ETA of the baseline nnUNet model?

Dear Giuseppe,

Sorry to hear that you have troubles accessing the ground truth labels. Our colleagues tried to access the data in incognito mode to make sure the page is not password-hidden, but that doesn't seems to be the problem. We recently set the labels to public, so maybe that could have been the issue. Could you maybe try it again or ask a colleague to access the page?

Let us know if this is not working out for you.

Kind regards,

Megan Schuurmans

Dear Ziteng,

Thank you for your message, happy to hear you were able to download the data! We hope to have the baseline model available within one or two weeks. I hope this will work out for your timeline as well!

Kind regards,

Megan Schuurmans

Hi,

When loading the data I saw that there were segmentations, but no images for the following study id's: {'100036_00001', '100433_00001', '101381_00001'}

I am not sure whether there is something going wrong on my side, but I cannot find those images in the batches I downloaded from zenodo.

Hi, automatic label 100051_00001.nii.gz has a 0 KB size, can you fix that?

Hi, if I understand correctly, the labels of Dataset103_PANORAMA_baseline_Pancreas_Segmentation should be binary masks of the pancreas. Are those labels different from Pancreas parenchyma (label 4)? If yes, will these labels be available as well?

Dear Thijs Gelton,

Thank you for your message regarding the segmentations of {'100036_00001', '100433_00001', '101381_00001'}. These images were used for the sanity check and, therefore, not provided for the training. As the sanity check does not contribute to the leaderboard position, you can ignore the masks of these cases. Our apologies for the confusion.

Kind regards,

Megan Schuurmans

Dear Ziteng Liu,

Thank you for your message. We will work on segmentation 100051_00001.nii.gz. Additionally, the labels provided are segmentations of the potential lesion and surrounding pancreas anatomy, including the pancreas parenchyma.

Kind regards,

Megan Schuurmans