The purpose of this challenge is to directly compare methods for the automatic segmentation of White Matter Hyperintensities (WMH) of presumed vascular origin.
To participate in the challenge, interested teams can register on this website. After registration, training data can be downloaded. This data consists of 60 sets of brain MR images (T1 and FLAIR) with manual annotations of WMH (binary masks) from three different institutes / scanners. Manual annotations have been made by experts in WMH scoring.
Participants can train their methods on the available data and submit the resulting method for evaluation by the organizers. A brief description (1-2 pages) of the method should be included, that optionally can be submitted for the LNCS post-proceedings. Test data will not be released, but consists of 110 sets of brain MR images from five different scanners (of which three are also in the training data).
Results & journal publication
After the deadline, all submitted methods will be evaluated by the organizers. During the challenge session at the MICCAI 2017, participants will present their methods (5 min oral pitch presentation and a poster). The final results will be presented by the organizers. All results will be summarized in a journal publication and each participating team who presented their method at the challenge session is allowed two co-authorships.
Programme 14 Sept
- Invited talk by Jerry Prince: “Dice overlap metrics for multiple objects: application to lesion segmentation”
- Power pitches by participants (2 min per team)
- Final results of the challenge + awards + discussion
- Posters during coffee break