Acknowledgments & Citation

The anonymous patient imaging data is being made available by the generous cooperation of Alexandra J. Golby, M.D, Principal Investigator of the Golby Lab in the Department of Neurosurgery at Brigham and Women's Hospital, Harvard Medical School. Dataset selection and processing was thanks to Isiah Norton. Thanks also to the San Diego Supercomputer Center for data storage and archiving.

We require that any publications (not the contest entries) using any of these patient datasets include the following acknowledgment:

Brain tumor imaging data courtesy of Alexandra J. Golby, M.D in the Department of Neurosurgery, Brigham and Women's Hospital, Harvard Medical School, supported by NIH grant U41-RR019703.
It would also be helpful to include the http://viscontest.sdsc.edu/2009/data/ URL for the datasets.

Dataset Content and Format

MRI data is acquired on a 3-dimensional regular grid, and the data is presented here in a format that handles multi-variate image data, Nrrd. As used here, this format is essentially just a human-readable ASCII header and a raw data block in a separate file.

The structural MRI scans are intended to provide a higher-resolution anatomical context that is available in the diffusion MRI. The faces have been volumetrically removed from the data to ensure patient anonymity.

For the diffusion MRI, Nrrd is used to store the multiple diffusion-weighed images that are acquired to capture the directional diffusion patterns at each image sample. This data is also in Nrrd format, as documented here. This is the format for dMRI used for surgical planning research at Brigham and Women's Hospital. In the Nrrd DWI documentation, pay especial attention to the "measurement frame" field, which serves the important role of describing how the coordinate frame of the diffusion gradients (and the diffusion tensor) relate to the coordinate frame of the image sampling grid itself.

Although contestants may choose to not use diffusion tensors to model the dMRI data, or may choose to estimate diffusion tensors differently, we make available volumes of diffusion tensors, estimated per-sample from the diffusion MRI data, also in Nrrd format, but with seven values per sample (according to the 3D-masked-symmetric-matrix per-axis sample kind). The seven values are:

  1. Brain mask (values are 0.0 outside brain to 1.0 inside, with a thin boundary region)
  2. Dxx
  3. Dxy
  4. Dxz
  5. Dyy
  6. Dyz
  7. Dzz

Synthetic data

To verify that the diffusion data is being read and processed correctly, it is helpful to have a non-trivial synthetic dataset for which there is a known correct tractography result. We have generated a synthetic dataset of a helix for this purpose, both as a DWI volume and as pre-computed tensors: The image at left is a rendering of some simple tractography results through this data. Notice that the primary helix is right-handed, and that the fibers follow the primary helix in a secondary helical path, also right-handed. Make sure that you can generate a comparable tractography result prior to working with the patient data.

Patient data

2009 contest has been deferred to 2010.

Click here to visit 2010 Vis Contest webpage

Patient A:
  • Structural: A/A-spgr-deface.{nhdr,raw}
  • Diffusion images: A/A-dwi.{nhdr,raw}
  • Diffusion tensors: A/A-ten-mask.{nhdr,raw}
  • Functional: coming soon
  • Tumor segmentation: coming soon
  • Affine Registration between datasets: coming soon
Patient B:
  • Structural: B/B-spgr-deface.{nhdr,raw}
  • Diffusion images: B/B-dwi.{nhdr,raw}
  • Diffusion tensors: B/B-ten-mask.{nhdr,raw}
  • Functional: coming soon
  • Tumor segmentation: coming soon
  • Affine Registration between datasets: coming soon
The datasets have the following checksum information (determined by cksum) 
1431380589	2094	A/A-dwi.nhdr
1101094629	159383552	A/A-dwi.raw
2160316750	3299	B/B-dwi.nhdr
1178940696	389021696	B/B-dwi.raw
2459858664	551	A/A-spgr-deface.nhdr
2748540511	131072000	A/A-spgr-deface.raw
857730749       549     B-spgr-deface.nhdr
3897384115	15728640	B/B-spgr-deface.raw
563830772	603	A/A-ten-mask.nhdr
2582920851	69730304	A/A-ten-mask.raw
3356613123	604	B/B-ten-mask.nhdr
3034390199	97255424	B/B-ten-mask.raw

Utility Software