Submit your results
Note: Uploading may take a while, press the button only once and wait for a new screen to appear!
- Each submission should be a single compressed archive (zip or rar) containing the segmentations of all images. Segmentation files should be directly in the root of the archive, and not nested in a folder structure. Each segmentation should be a MHD/RAW file (e.g. Segm_MRBrainS13_01.mhd and Segm_MRBrainS13_01.raw) or a Nifti file (e.g. Segm_MRBrainS13_01.nii) of type 8 bit unsigned char. Read the Details page for more information.
The resolution and dimensions of the segmentation results should be the same as the thick-slice T1-weighted scan (voxel size: 0.958mm x 0.958mm x 3.0mm).
- Results should be named Segm_MRBrainS13_01 to Segm_MRBrainS13_15. Within these files the segmented tissue should be labeled as follows:
- Background (everything outside the brain)
- Cerebrospinal fluid (including ventricles)
- Gray matter (cortical gray matter and basal ganglia)
- White matter (including white matter lesions)
With each submission, a short description of the segmentation algorithm (1-2 pages) should be provided. Multiple submissions are allowed, as long as significantly different approaches are used for each submission, the difference should be described in the method description. Updating an existing submission is allowed for minor tweaks to an algorithm up to three times (if you want to update an existing submission, please sent an e-mail to email@example.com).
Please consider the following guidelines for the content of the method description:
- Is your algorithm automatic or semi-automatic? Describe the required user input, and the average time spent per scan, for semi-automatic algorithms.
- Which sequences are used by your algorithm? Only the thick-slice T1-weighted scan, or the IR, FLAIR and thin-slice T1 scan as well?
- List the overall structure of the algorithm in a step-wise fashion and describe each step of the algorithm in detail. Include pre- or post-processing steps, when required.
- Mention limitations of the algorithm. Was the algorithm specifically designed to segment only certain types of scans (e.g. scans of healthy volunteers or patients with specific pathology)? Was it optimized to work for scans with thick or thin slices? Was your algorithm specifically designed for a certain goal (e.g. volume measurement to quantify atrophy in patients with Alzheimer's disease)? Etc.
- Was the algorithm trained with example data other than the training data provided for the MRBrainS challenge? If so, describe the characteristics of the training data.
- Which labeled structures were used to fine-tune your algorithm (e.g. cortical gray matter and basal ganglia were used seperately or combined labeled as gray matter)?
- If the algorithm has been tested on other databases, you could consider including those results.
- What is the average runtime of your algorithm, and on which system is this runtime achieved?