Dear Organizers of the BraTS Challenge,
I am encountering problems in understanding what metrics I will have to include in the short paper for task 8.
In the task description, the stated scores are:
- Healthy, Tumor SSIM
- Dice Scores (3 tumor structures)
- Hausdorff Distance (3 tumor structures)
While in the table of the evaluation phase I can see:
- Dice scores (4 tumor structures, recession cavity too?????)
- Normalized Surface Distance (4 tumor structures again)
What are the right metrics for that task?
Created by OMAR CARPENTIERO ocarpentiero hi Omar @ocarpentiero,
The two subsets, GLI and METS, share three structures of interest: Enhancing Tumor (ET), Tumor Core (TC), and Whole Tumor (WT). Therefore, metrics related to ET, TC, and WT will be used for the final ranking. Please report these metrics in your method description.
For segmentation evaluation, we are using Dice scores (overlap-based) and Normalized Surface Distance (distance-based) across all segmentation tasks this year (Last year, we used Hausdorff Distance instead). Apologies that this update hasn't yet been reflected on the website.
Best regards,
Hongwei (Bran)