Dear @BraTS2024Participants,
We are very excited to present the 2024 BraTS Meningioma Radiotherapy Challenge. Please use this thread for any direct communications with me and @dlabella29, the co-organizers of this challenge. We will also post important updates about this challenge here.
Here are some of the unique aspects of the BraTS-MEN-RT challenge that you should know about:
1) Image data consists of a a single series (3D postcontrast T1-weighted spoiled gradient echo imaging). This is the only sequence required for meningioma radiotherapy planning.
2) All image data is in original, native acquisition space, and therefore the image matrix and voxel resolution varies from scan to scan. This reflects real-world MRI data!
3) Image data have undergone defacing rather than skull stripping as used in other BraTS challenges. This makes the data closer to real-world MRI data, and also avoids cropping meningiomas that extend outside the skull.
4) The segmentation target consists of a single label, which corresponds to the gross tumor volume (GTV) that is used for radiotherapy planning.
5) BraTS-MEN-RT participants will be ranked based on global Dice score, which differs from some of the other challenges that are using lesion-wise Dice for ranking.
Please let us know if you have any questions, and best of luck with BraTS-MEN-RT!
-Evan Calabrese
BraTS-MEN-RT co-organizer
Created by Evan Calabrese ecalabr I believe you are correct, but lets get Rachit to give us the details of implementation. @rs2492 @rsaluja Dear @ecalabr,
Thank you for your answer. So, it does not penalise false positives but penalises false negatives like the lesion-wise metric, i.e., if there is a false negative and and true positive the dice score will be divided by two?
Thank you! Thanks for your question. We have had much internal debate over this. We originally opted for global dice because we did not want to penalize false positives, as these can easily be removed by a treating radiation oncologist and not all meningioma lesions will always be treated at a given radiotherapy session. However, the larger BraTS organizing team has expressed significant concerns regarding global Dice and its strong dependence on lesion size.
We have eventually settled on a version of lesionwise dice that does NOT penalize false positives. Note that this is more similar to global Dice compared to the lesionwise Dice metric used in other BraTS 2024 challenges. @rs2492 @rsaluja can you please provide more specifics for our participants? Thanks!
-Evan Calabrese Dear organisers,
I mention that "participants will be ranked based on global Dice score", however, in the validation table, the results are the Lesion-wise.
Is there a mistake?
Thank you!
Drop files to upload
??[Task Specific Discussion] BraTS-MEN-RT Challenge 2024 ??? page is loading…