By: kamilrzechowski on June 20, 2024, 7:39 p.m.
Hi. I have a question about the time to biochemical recurrence. The minimum time for the follow-up years is 0.030116359 in the given dataset and it is for the even 0 (no biochemical recurrence. case_radboud_0182). Isn't it too short to confirm no biochemical recurrence for the given case? There are 138 cases with the event 0, that have less than 2 years followup.
By: KhrystynaFaryna on July 15, 2024, 2:50 p.m.
Dear participant,
Thank you for your question. We understand that the time to follow of less than 2 years is too short to confirm no recurrence. The main evaluation metric of this challenge is a censored C-index, where all non-event cases are censored. For the model development, we leave it up to participants to determine the censoring cut-off. Whether the cases with short follow-up time would be useful for the model development very much depends on the approach one follows - less data/less noise vs. more data/more noise tradeoff. In the FAQ we referenced several papers that use deep learning to predict prostate cancer recurrence, where participants can see common thresholds to use(e.g. 4 or 5 years).
I hope this helps. Do not hesitate to reach out in case of other questions We are looking forward to your submission.
Kind regards,
Khrystyna
