Spaces for imaging data

This section describes the spaces used in the prepared NSD data. Understanding this information is important for appropriate handling of the imaging data.

Spaces for the pre-processed data


Each subject has two functional data preparations: “func1mm” and “func1pt8mm”. This refers to either preparing the data at 1-mm spacing (i.e. upsampling the data) or at 1.8-mm spacing. The 1-mm data has additional fine-scale detail, but is very large in size (approximately 6 times larger than the 1.8-mm data). There is also a difference in temporal resolution in the pre-processed data: the temporal sampling rate (TR) for the two preparations is 1 s and 1.333 s, respectively.
 
The two functional data preparations are in the same physical space. For example, the two preparations share a common first “corner” voxel (located at anterior, right, inferior) and the data from this voxel are identical across the two preparations. However, the two preparations have slightly different fields-of-view (since the voxel sizes do not evenly divide).
 
Each subject has three anatomical data preparations: “anat0pt5”, “anat0pt8”, “anat1pt0”. This refers to preparing the anatomical data (e.g. T1, T2) at 0.5-mm, 0.8-mm, and 1.0-mm resolution. All three versions share exactly the same field-of-view and are centered at exactly the same location in space.
 
The functional and anatomical data are not in register; rather, we have estimated a nonlinear warping for each subject that specifies how the functional data can be registered to the anatomical data (and vice versa). In some cases, we provide convenient versions of the data that have already been mapped (e.g. a version of the T1 that is warped and matched to the functional data).
 
There are three other spaces of note:
  • Some analysis results are prepared in FreeSurfer's surface space, and they are either contained within FreeSurfer directories (e.g. "label") or in directories named nativesurface.
  • Some analysis results are prepared in MNI space. This is achieved based on a nonlinear warp estimated for each subject that takes their 1.0-mm T1 and matches a 1-mm MNI template.
  • Some analysis results are prepared in FreeSurfer's fsaverage space. This is achieved based on the curvature-based alignment provided by FreeSurfer.

Note that for the diffusion data, the pre-processed volumes are matched to the anat0pt8 space (0.8-mm).

Basic handling of NSD data files


All NIFTI files in the prepared NSD data are in LPI ordering (the first voxel is Left, Posterior, and Inferior). In addition, all NIFTI files have their origin set to the exact center of the image slab, with one exception being NIFTI files in MNI space (for details, see  Technical notes ).

We have pre-computed transformations that map between the various spaces, and these transformations are incorporated into the lightweight utility nsd_mapdata. This utility transforms user-supplied data from one space to another using interpolation (see  Code  for details).