Submission details, baseline container and datacentric code

Dear all,

The baseline algorithms are now available at https://github.com/ClinicalDataScience/autoPETIII. If you encounter any issues, please raise an issue in the repository. Detailed instructions on algorithm containerization and submission are also provided here. Happy testing!

The datacentric code got also a little update (you can now use the option fold: all) and a small bugfix for the fpv metric.

Best AutoPET III Team

Dear Organizer,

I would also like to inquire about the number of submissions I can make for the preliminary test phase. Since the system says I can make 5 submissions for the next 7 days, does that mean I can make some further submissions after that 7 days? Also, do you count all submissions using the same algorithm once (since we can update the algorithm), or do you count them separately? Thank you very much!

Yours sincerely, Lennon

Dear Lennon,

The counting mechanism of GC only tracks successful submissions, and you can make up to 5 successful submissions per week. This means you can submit more next week if needed. Submissions are independent of the algorithms used. Generally speaking, the preliminary submission is intended for testing the Docker setup (files are drawn from the training set). For the final submission, you only have two attempts, so everything must be fully functional. Repeatedly pushing the same Docker image while only changing the weights won't provide new insights. This limit is in place due to limited computational resources. I hope this clarifies your question.

Best Jakob

Dear Jacob,

Thank you very much!

Yours sincerely, Lennon

Dear Organizers,

I have another question regarding the submission. It was written in Final Test Submission description that we should "Publish a brief method description and results on ArXiv or a similar platform, linked to your final submission", however, we are only able to upload a PDF file under "Preprint paper*". Could you please clarify if we should also post a link to the ArXiv (or similar) preprint publication and whether it should be public by the submission deadline or whether it is enough to upload our PDF file?

Thank you!

Dear Shuhan,

yes the preprint should be published on ArXiv (or similar) and publicly available. In the submission tab you could just upload the published PDF (with the timestamp visible). If no timestamp or similar is visible on the PDF we will search for the paper latest on Tuesday 17.09. (based on arXiv's announcement schedule).

Best

Jakob

Dear Jacob, I have used vixra as alternative to ArXiv does not provide a visible timesamp on a pdf. https://vixra.org/faq.html

No problem, as long as everything is publicly available :)

Dear Organizer,

It has come to my notice that we need to make our algorithms publicly available as per the rules of the challenge:

"Algorithms must be made publicly available as e.g. GitHub repository with a permissive open source license (Apache Licence 2.0, MIT Licence, GNU GPLv3, GNU AGPLv3, Mozilla Public Licence 2.0, Boost Software Licence 1.0, The Unlicence)"

But I'd like to ask whether that means we only need to make our weights publicly available (i.e. our modification of https://github.com/ClinicalDataScience/autoPETIII) or do we need to make our entire project including the codes we used to train our model available? Thank you very much!

Yours sincerely, Lennon

Dear Lennon,

see here for further clarification https://grand-challenge.org/forums/forum/automated-lesion-segmentation-in-whole-body-petct-multitracer-720/topic/extended-final-submission-and-paper-deadline-further-clarification-2737/

Best Jakob

Dear Jakob,

Thank you very much!

Yours, Lennon