Announcements¶
August 2022
The preprint of the summary paper of the challenge is available at https://arxiv.org/abs/2208.07167
Sudre, C. H., Van Wijnen, K., Dubost, F., Adams, H., Atkinson, D., Barkhof, F., ... & de Bruijne, M. (2022). Where is VALDO? VAscular Lesions Detection and segmentatiOn challenge at MICCAI 2021. arXiv preprint arXiv:2208.07167.
Please do refer to it and cite it when using material from the challenge!
October 2021
Currently submission is closed. In the near future we will have a live leaderboard and submitting will be possible again¶
August 2021
The final deadline is coming up!¶
It's time to submit your final method on August 5th! Please check the submission page for how to submit and the docker page on how to prepare your docker with our template. Please be sure to do the following for your final submission:
- Submit your Docker as *.tar.gz
- Submit the corresponding Dockerfile
- Fill in the submission questionnaire of the task(s) you are participating in
- Upload a pdf of 1 to 2 page describing your method and your initial results (you will be asked to upload this in the submission questionnaire)
Some important remarks:
- If your docker requires a GPU to run, please make sure your Docker is compatible with the Ampere GPU architecture (or reach out to us if this is an issue). See the submission page (section further important remarks) for more information
- We expect all methods to be submitted on August 5th, but we will have a grace period until August 10th. On the 10th we need to have all requested files in order for your submission to be valid.
- We will run your Docker on an example (a training set case) and send it to you so that you can check it is what you expected. Please keep an eye out for this email and let us know as soon as you can if your method is working as it should or if there are any discrepancies.
- Please be sure to change the labels in the Dockerfile according to what your method needs! Especially the required number of cpu cores, number of gpus, gpu memory and RAM (if your method does not use a GPU please fill in 0 for the number of gpus label)
July 2021
2nd validation period starts soon!¶
From Monday July 12th until Monday July 19th you can submit your method to test against the validation set (a small separate set). Please see the submission page to see details on how to submit and the docker page on how to prepare your docker with our template!
- Note that we adapted our template since the first validation period.
- If your docker requires a GPU to run, please make sure your Docker is compatible with the Ampere GPU architecture (or reach out to us if this is an issue). See the submission page (section further important remarks) for more information
June 2021
Validation period is on!¶
Until Monday 21st June you can submit a first version of your submission that will be assessed against a small set. The second validation period will open on the 12th July for one week. Please see the submission page to see details on how to submit!
April 2021¶
The updated training data is now available on Zenodo!¶
The link you have been given at registration will bring you to the Zenodo page where you can download this new version. First click on the most recent version number on the right side and then download the data. The following changes have been done in this update:
- The region masks have been updated for the RSS cases. The region masks have been expanded slightly to cover the annotated areas better. Additionally, the labels per region are now consistent across all region masks (see the page with extra information on Task 1 for more information) (Task 1: enlarged PVS)
- Two versions of the SABRE segmentation masks are now available. One version containing the segmentations in slabs as originally intended. The other version contains segmentations for the full brain like in the previous version of the dataset (Task 1: enlarged PVS)
- The count files for the RSS segmentation cases have been updated and now contain the correct counts and correspond to the consistent region labels. The count files for all cases now contain the correct counts (Task 1: enlarged PVS)
- The segmentation masks for SABRE case 104 and RSS case 222 have been added (Task 2: microbleeds)
- The range of the segmentation masks have been updated: 0 for background and 1 for microbleeds (Task 2: microbleeds)
- The slight misalignment between T2* scans and segmentation masks for ALFA cases has been fixed (Task 2: microbleeds)
- 'sub-105_space-T1_possible.nii.gz' has been removed (Task 3: lacunes)
April 2021¶
We will soon release an update of the training data, in which the following issues will be fixed:¶
- The SABRE segmentation masks should only contain segmentations for slabs, but currently also contain segmentations outside the slabs indicated in the region masks. In the updated dataset we will add the originally intended version of the segmentation masks (so segmentations in slabs = current segmentation masks * corresponding region mask), we will also keep the current version of the segmentation masks (Task 1: enlarged PVS)
- Range of RSS region masks should be 0 - 6, currently this is 0-7 for some cases due to overlapping regions. Region labels are not consistent between masks. The region labels will be consistent in the updated dataset (Task 1: enlarged PVS)
- The count files for the RSS segmentation cases contain incorrect counts for some of the regions. The count files for the RSS count cases (the cases with only a count file, no segmentation masks) do contain correct counts for all regions. We will provide corrected count files for the RSS segmentation cases (Task 1: enlarged PVS)
- Missing segmentation masks for SABRE case 104 and RSS case 222, these will be added in the updated dataset (Task 2: microbleeds)
- Range of RSS segmentation masks should be 0 for background and 1 for microbleeds, currently this is 32768 for background, 32769 for microbleeds (Task 2: microbleeds)
- Range of ALFA case 301 should be 0 for background and 1 for microbleeds, currently this is 0 for background, 2 for microbleeds (Task 2: microbleeds)
- ALFA cases have a slight misalignment between T2* scans and segmentation masks, we will align the segmentation masks with the T2* scans (Task 2: microbleeds)
- SABRE case 105 contains a file 'sub-105_space-T1_possible.nii.gz' which can be ignored and will be removed in the updated dats (Task 3: lacunes)