Welcome to Discourse!


Thanks for joining us on the official discussion forum of the 2019 KiTS Challenge! Please utilize this space for any questions, comments, or concerns you may have about the data and the challenge.

Edit: I think my wording was unclear - by “this space” I mean the forum as a whole. Please feel free to start new topics outside of this thread!

pinned #2


Hi there, Can anyone guide me to sample datasets for working before the official ones are out?


I would recommend that you take a look at the data from the LiTS Challenge. Their task is also segmentation in abdominal CT.

The LiTS data was released in the NIFTI format (.nii) where ours will be in DICOM, but don’t worry too much about loading the raw data. We will provide some starter code (python) to help with preparing the data for training.


How often can we submit the result?


Submission will open 2 weeks before the deadline. You can submit as often as you like during that time, but scores will not be shown until the submission period is closed.

Clarification requested about multiple submissions

We think it would be better if we can submit 1 or 2 results per data and know the scores immediately. The reason is if we know the scores, we can know how well we have done and where we can improve our methods.


Thank you for the suggestion. Our intention when designing this challenge was to adhere as closely as possible to recent literature about best practices in challenge design and to the design of past challenges that were successful. The relevant recommendation of Reinke et al. from their MICCAI 2018 paper on challenges [1] is summarized in Table 2: “Feedback after a submission should not reveal information on individual cases”. Further, the highly successful ISIC challenge last year did as we are proposing to do, and did not make any results available until after the deadline.

Shortly after the August 1, 2019 MICCAI deadline has passed, the challenge will reopen for submissions indefinitely and the leaderboard will then update in real time.

[1] Reinke, Annika, et al. “How to exploit weaknesses in biomedical challenge design and organization.” International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, Cham, 2018.


Sir can we use any kind image segmentation algorithms to return the results? I mean is it necessary to use neural network based algorithm to segment the image?


Of course! As long as a method does not include manual intervention, it’s allowed.


when will you release training dataset ?


You can expect it sometime within the next 12 hours. I will make a new Discourse topic and update the challenge website with instructions once the data is available.


Just waiting on compression and then an upload at this point. I can’t imagine that it will be more than a couple of hours longer, sorry for the delay.