Processing of the Human Connectome Project data with the CBS high-res Brain Processing Tools

This folder contains the results of our processing of the data made freely available through the Human Connectome Project (http://humanconnectome.org) [HCP]. It includes segmentations of the whole brain and cortical surfaces for the cerebrum (left/right separated) and cerebellum at the original resolution of 0.7 mm isotropic performed with the CBS Tools (http://www.nitrc.org/projects/cbs-tools/). The original data is governed by the HCP Opean Access data use terms, and the processed data is provided "as is", with no guarantee of fitness for any specific purpose. The processing follows the pipeline outlined in this publication [CBS-Tools] with modified pre-processing and additional currently unpublished developments, in particular for the cerebellum.


The processed data is organized as follows.

In "preprocessing", the T1-weighted and T2-weighted images have been reoriented to RAI orientation, skull-stripped with SPECTRE and normalized. An additional filter image locating regions of dura mater, csf partial voluming and blood vessels is also included.

In "whole_brain", a segmentation of the brain into its main cortical and sub-cortical regions in provided. The segmentation includes a labeling and a MGDM distance function, which can be used to reconstruct a level-set surface representation of any structure or group of structures. Strict topology constraints guarantee homeomorphism of single and multiple structures (see [MGDM] for details).

In "cerebrum_left", "cerebrum_right" and "cerebellum", cortical segmentations for the GM/WM boundary ("gwb"), CSF/GM boundary ("cgb") and an average surface inside the cortex ("avg") are given as level-set images (NIFTI format) and 3D meshes (VTK format). An additional mesh of inflated average surface ("inf") is also included, as well as voxel-level segmentations of the cortex ("cortex") for the three structures. Segmentations were performed with the CRUISE algorithm and the cortical inflation with the partial inflation algorithm from [Surf-Norm].


Formatting details.

The volumetric data is stored in NIFTI format, although the CBS Tools work in voxel space and the NIFTI orientation matrices are ignored in the processing. The 3D meshes are encoded in MIPAV coordinates, which are give as follows:
x_mesh = x_voxel * rx
y_mesh = ( (ny-1) - y_voxel ) * ry
z_mesh = ( (nz-1) - z_voxel ) * rz
Level-set surface representations are volumetric images of the signed distance to the surface (negative inside, distances are given in voxels). They are bounded to arbitrary values far away from the surface.


References:

[HCP] The WU-Minn Human Connectome Project: An overview. 
Van Essen DC, Smith SM, Barch DM, Behrens TEJ, Yacoub E, Ugurbil K, for the WU-Minn HCP Consortium.
NeuroImage 80(2013):62-79. 
http://www.ncbi.nlm.nih.gov/pubmed/22366334

[CBS-Tools] A computational framework for ultra-high resolution cortical segmentation at 7Tesla.
Bazin PL, Weiss M, Dinse J, Schäfer A, Trampel R, Turner R.
Neuroimage. 2013 Apr 25. pii: S1053-8119(13)00332-7. doi: 10.1016/j.neuroimage.2013.03.077.
http://www.ncbi.nlm.nih.gov/pubmed/23623972

[SPECTRE] Simple paradigm for extra-cerebral tissue removal: algorithm and analysis.
Carass A, Cuzzocreo J, Wheeler MB, Bazin PL, Resnick SM, Prince JL.
Neuroimage. 2011 Jun 15;56(4):1982-92. doi: 10.1016/j.neuroimage.2011.03.045.
http://www.ncbi.nlm.nih.gov/pubmed/21458576

[MGDM] A Multiple Object Geometric Deformable Model for Image Segmentation.
Bogovic JA, Prince JL, Bazin PL. Comput Vis Image Underst. 2013 Feb 1;117(2):145-157.
http://www.ncbi.nlm.nih.gov/pubmed/23316110

[CRUISE] CRUISE: cortical reconstruction using implicit surface evolution.
Han X, Pham DL, Tosun D, Rettmann ME, Xu C, Prince JL. Neuroimage. 2004 Nov;23(3):997-1012.
http://www.ncbi.nlm.nih.gov/pubmed/15528100

[Surf-Norm] A geometry-driven optical flow warping for spatial normalization of cortical surfaces. 
Tosun D, Prince JL. IEEE Trans Med Imaging. 2008 Dec;27(12):1739-53. doi: 10.1109/TMI.2008.925080.
http://www.ncbi.nlm.nih.gov/pubmed/19033090