Image data used in this challenge were acquired from five different scanners from three different vendors in three different hospital in the Netherlands and Singapore. For each subject, a 3D T1-weighted image and a 2D multi-slice FLAIR image are provided. The manual reference standard is defined on the FLAIR image.
|UMC Utrecht||3 T Philips Achieva||20||30|
|NUHS Singapore||3 T Siemens TrioTim||20||30|
|VU Amsterdam||3 T GE Signa HDxt||20||30|
|3 T Philips Ingenuity||0||10|
|1.5 T GE Signa HDxt||0||10|
For each subject, the following files are provided:
|/orig/3DT1.nii.gz||The original 3D T1 image, with the face removed|
|/orig/3DT1_mask.nii.gz||Binary mask used to remove the face.|
|/orig/FLAIR.nii.gz||The original FLAIR image. This image is used to manually delineate the WMH and participants should provide the results in this image space.|
|/orig/reg_3DT1_to_FLAIR.txt||Transformation parameters used to align the 3D T1 image with the FLAIR image. Participants can use this to do the transformation themselves, in case they use the 3D T1 for processing. Results will be evaluated in the FLAIR space. Call: transformix -in IMAGE.EXT -out /output -tp /orig/reg_3DT1_to_T2_FLAIR.txt|
|/orig/T1.nii.gz||The 3D T1 image aligned with the FLAIR image.|
|/pre/3DT1.nii.gz||Bias field corrected 3D T1 image.|
|/pre/FLAIR.nii.gz||Bias field corrected FLAIR image.|
|/pre/T1.nii.gz||Bias field corrected T1 image, aligned with FLAIR.|
|/wmh.nii.gz||Manual reference standard, only available for training data.|
This folder contains the original images, anonimized and having the face removed. The mask used to remove the face from the 3D T1 images is provided. The 3D T1 image has been aligned with the FLAIR image using elastix 4.8 with the following parameter file: Elastix parameter file for the WMH Segmentation Challenge.
All images were pre-processed with SPM12 r6685 to correct bias field inhomogeneities.
This file is only provided for the training data. It contains the following three labels:
- White matter hyperintensities
- Other pathology
The objective of this challenge is to automatically segment WMH. Because we do not require methods to identify all other types of pathology, we provide a rough mask for them that will be ignored during evaluation.
Participants are allowed to use additional data to train their method. This must be mentioned in the description that is submitted with the method. We encourage participants to use open data or make the additional data open access.
Participants who have registered their team and signed the confidentiality agreement are eligible to download the data.