Training data¶
This page contains all the material for the training data.
Diffusion weighted images and sampling schemes¶
The diffusion weighted images have been generated for 3 predetemined sampling schemes, accordingly to the sampling classes of the contest (see presentation of sampling classes). If you have an optimized sequence or specific needs for the acquisition, you are encouraged to request additional diffusion weighted images of the same training data. Simply send an email to the organisers with a .bval (table of b-values) and a .bvec (table of gradient directions) file. Be careful that we limit to a maximum of 5 the number of requests per participating team.
| Description | Sampling scheme | Diffusion-weighted images |
|---|---|---|
| 32 directions, b = 1200 s/mm² | ||
| 64 directions, b = 3000 s/mm² | ||
| 515 acquisitions, b < 12000 s/mm² |
Mask and seeding regions¶
The file mask.nii.gz is the binary mask for the groud-truth. The seeding regions for tracking are stored in a file rois.nii.gz. They correspond to the ends of the fiber bundles. More specifically, fiber 1 connects region labelled 1 to region labelled 2; fiber 2 connects region labelled 3 to region labelled 4, and so forth.
Ground-truth fiber geometries¶
Fiber geometries are provided in a text format. Each file stores the trajectory of a specific fiber. The fiber bundles radii are also provided in a separate file.