By: Yeah on March 12, 2021, 2:25 a.m.
After reading the documents and data, I can't understand how to map the data of Task 1. Could you please explain it in more detail? Also, the categories of data can be listed. I found 0-7 in task 1, and 0-3 in task 2. But I don't know what they mean. Is it a binary segmentation problem?
By: kvanwijnen on March 26, 2021, 8:54 p.m.
Thank you for your message and great to hear you are participating in our challenge!
For task 1 region masks indicate which parts of the brain have been annotated by a rater (for task 2 and 3 the full scans have been annotated). It is indeed a binary segmentation problem, the submitted methods should output a binary segmentation mask for the full brain. The training set contains cases with counts and cases with segmentations, so participants can choose which annotation type and which cases they want to use to develop their method. The segmentation masks are indeed binary, the background is 0 and the foreground is 1. Counts are provided in csv files.
You refer to categories of 0-7 in task 1, I believe you are refering to the region masks, these are only used to indicate which parts of the brain have been annotated, so this is not what your method should output. We added more details on this on this page: https://valdo.grand-challenge.org/Task1/. The desired output for this task is a binary segmentation mask such as is given for 12 of the 40 cases (called 'PVSSeg'), however the methods should output a segmentation mask for the whole brain (these masks only contain segmentations for part of the scans and the method should output segmentations for the full brain).
For task 2 the microbleed segmentation masks (called 'CMB') should be binary as well, so 0 for BG, 1 for microbleeds. There shouldn't be values of 2 or 3 in the segmentation masks. For which cases did you encounter a value of 2 or 3 in the CMB nifti? It would be great if you could let us know!
