Data

[Madeypo]

Dear Mart,

In the README, the data/example.ipynb in HookNet-TLS Annotations: https://zenodo.org/records/10614928 only converts XML to PNG, which may not be the Tissue-Background Segmentation Algorithm as in the Grand Challenge. How can I obtain the Segmentation Algorithm since the application is still pending?

Regarding the HookNet-TLS Annotations Masks: https://zenodo.org/records/10635034, the TIF files cannot be opened and show as damaged. This should not be due to my computer, as other TIF files can be opened normally. Even after re-downloading the masks, it does not work.

Additionally, I would like to ask if it is possible to successfully deploy the model locally since it can be deployed in Docker.

Thank you very much.

[martvanrijthoven]

Dear Madeypo,

Unfortunately i am not the owner of the tissue background segmentation algorithm. I would advice you to contact grand-challenge support. Alternative you could use something like Otsu's method to generate a tissue mask.

The tiff files can be opened as an overlay with ASAP software (https://computationalpathologygroup.github.io/ASAP/) Furthermore you can openen them in python with wholeslidedata (https://github.com/DIAGNijmegen/pathology-whole-slide-data) and the "asap" image backend. Please let me know if you still have trouble opening them.

Yes the model can be deployed locally. Please le me know if you have any further questions about it.

Best wishes, Mart

[Madeypo]

Dear Mart,

How can I generate a WSI-sized mask image from XML? This might be limited by Python libraries, making it impossible to obtain high-resolution mask images.

Best wishes

[martvanrijthoven]

Dear Madeypo,

Here is an example to generate WSI sized mask from xml annotations files: https://github.com/DIAGNijmegen/pathology-whole-slide-data/blob/main/scripts/convert_annotations_to_mask.sh

However for HookNet-TLS annotations i already created them and you can download them here: https://zenodo.org/records/10635034