DWI Files

Overview

The following folders and files are produced by the dtiInit analysis. The input files are n the left: the t1.nii.gz and the dwi files. The output files dtiInit are in the folders on the right. Detailed information about the files and images is below.

          ______________ <subdir> ____________
          |          |                 |
      t1.nii.gz    <ROIS>      <dti|NumDirs|InterpMethod>
      dwi.nii.gz                       |
      dwi.bval                _________|_______________________________________
      dwi.bvec               |         |          |           |               |
     Xform files            <bin>     dt6.mat   t1pdd.png    dtiInit-log.mat   <fibers>
                             |
                           tensors.nii.gz
                           b0.nii.gz
                           brainMask.nii.gz
                           vectorRGB.nii.gz
                           wmMask.nii.gz
                           wmProb.nii.gz
                           faStd.nii.gz
                           mdStd.nii.gz
                           pddDispersion.nii.gz

Input (root folder)

• T1 anatomical volume that can serve use as a background file. If you don't have a T1, we substitute in the MNI template.
• The diffusion weighted data (nifti, bvecs, and bvals). Some people put these data in a 'raw' sub-directory (not shown here)
• ROIS - A directory where regions of interest are stored. These are used for some types of fiber tracking, but there are no requirements

Some files representing transform coordinates may be found here. For example, the Xform files for acpc coordinates and for the eddy current correction can be stored here. (Not sure why).

Output folder

The name of the output folder describes two parameters of the analysis:

• numDirs - the number of diffusion directions measured
• InterpMethod - the interpolation method

Examples are dti31trilin or dti60trilinrt (robust tensor fitting). In a typical analysis, we might align the DWI data to the b=0 measurements using a trilinear interpolation method with eddy current correction turned on.

Top level output files

• dt6.mat - This file is mainly used by mrDiffusion, the DWI -user interface that is part of vistasoft
• dtiInit-log.mat - The parameters submitted to dtiInit are stored here. There are also data for reproducibility (various shah codes, versions, and so forth).
• t1pdd.png - A T1 background (either yours or the MNI template if you didn't have one) and the principal diffusion directions as an RGB overlay

The bin directory

We store calculated quantities in this directory.

All of the data in this directory are in the 'aligned' and 'interpolated' coordinate frame. Usually they are also cropped to a restricted region around the brain. The cropping changes the from the original data.

• tensors.nii.gz - Estimated diffusion tensors (different methods for computing, why we compute them)
• b0.nii.gz - The mean of the b=0 measurements. These data are resampled and interpolated the same way the data are interpolated. alignment to the T1 anatomical (I think; but does this mean they are not aligned to the T1? Or is this after the raw B=0 are aligned?)
• brainMask.nii.gz - Binary mask of the brain locations (mrAnatExtractBrain.m, which uses FSL's BET)
• wmMask - Binary mask of the white matter locations
• vectorRGB - Principal diffusion direction data as a color map
• wmProb.nii.gz - Not sure, but its range is 0,255 for some reason
• faStd - Fractional anisotropy values at each brain location
• mdStd - Mean diffusivity values at each brain location
• pddDispersion -

Fibers directory

The fibers directory is where we store, um, fibers. The directory is created, but not always used.

Background images

The b0 image is the non-diffusion-weighted image. It is an image acquired with a b-value (diffusion weighting strength) of 0.

Mean diffusivity (MD) is the mean of the eigenvalues of the diffusion tensor or, equivalently, the trace of the diffusion tensor divided by 3. This parameter quantifies the size of the diffusion ellipsoid. The MD measure is invariant with respect to the orientation of the diffusion tensor and tends to vary little across brain tissue.

Fractional anisotropy (FA) is a measure of the degree of anisotropy in a tensor model of diffusion at a given voxel. FA values are bounded between zero (a perfect shape) and one (an infinitely long cigar shape).
''In the FA map FA values closer to one appear brighter.'' [http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=17379499&dopt=Abstract ''(Ref.)'']

The Vector RGB color map is created from FA values and the three vector elements. The three colors in this map correspond to the direction of the largest eigenvalue of the diffusion tensor (principle diffusion direction) with proportional intensity according to the FA.
''RED = left/right, BLUE = superior/inferior, and GREEN = anterior/posterior.'' [http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=17379499&dopt=Abstract ''(Ref.)'']

The T1 weighted image provides structural contrast for anatomical images of the brain. The relative signal intensity of voxels within the image depends upon the T1 value of the tissue in that voxel.
''Tissues that have a smaller T1 value appear brighter than those with larger values:''
''(White matter T1 value = ~600ms. Gray matter T1 value = ~900ms. Cerebrospinal fluid T1 value = ~4000ms) - all values for Field strength of 1.5T''