open-discussion
open-discussion > dti data
Jul 12, 2012 05:07 PM | Torsten Rohlfing
dti data
The FA file is indeed "fractional anisotropy", but in order to
store the values as integers, not floats, we multiplied all values
by 1000. So 943 is really FA=0.943.
The md, l1, and lt values are exactly what you are saying.
Now for the meaning of the eigenvalues, l1 through l3 - these do
NOT define any coordinate system. The eigen*vectors* define such
coordinate system for each tensor, but you can not derive these
from the eigenvalues.
You can get the eigenvector data for the dominant eigenvalue from
two additional, archives that we put for download on the NITRC
download page, http://www.nitrc.org/frs/?group_id=214
sri24_tensors_nifti.zip
sri24_tensors_nrrd.zip
The first has the data in NIFTI format, the second in NRRD.
For the remaining two eigenvectors, I currently do not have
the files ready. I can probably put these together also, but
it will take a little while.
Hope this helps.
Torsten
On 07/12/2012 05:08 AM, wrote:
The file "fa" contains the data of the fractional anisotropy and I
would expect the values of "fa" to be in the range [0,1]. When I
opened the file I figured out that the values ranged from 0 to
943. Does "fa" denote something else and not fractional
anisotropy?
According to your paper published in Human Brain Mapping, "md" is
the average of the eigenvalues of the diffusion tensor and "l1" is
the largest of the eigenvalues. Is "lt" the average of the other
two eigenvalues, i.e., (l2+l3)/2 ?
Do the eigenvalues l1, l2 and l3 are the values in the x, y, and z
co-ordinates of a global Cartesian system, or every set of (l1,
l2, l3) corresponds to a local orthogonal system of the
eigenvectors of the diffusion tensor? If they correspond to a
local orthogonal system is it possible for me to get the
eigenvectors for each eigenvalue?
store the values as integers, not floats, we multiplied all values
by 1000. So 943 is really FA=0.943.
The md, l1, and lt values are exactly what you are saying.
Now for the meaning of the eigenvalues, l1 through l3 - these do
NOT define any coordinate system. The eigen*vectors* define such
coordinate system for each tensor, but you can not derive these
from the eigenvalues.
You can get the eigenvector data for the dominant eigenvalue from
two additional, archives that we put for download on the NITRC
download page, http://www.nitrc.org/frs/?group_id=214
sri24_tensors_nifti.zip
sri24_tensors_nrrd.zip
The first has the data in NIFTI format, the second in NRRD.
For the remaining two eigenvectors, I currently do not have
the files ready. I can probably put these together also, but
it will take a little while.
Hope this helps.
Torsten
On 07/12/2012 05:08 AM, wrote:
The file "fa" contains the data of the fractional anisotropy and I
would expect the values of "fa" to be in the range [0,1]. When I
opened the file I figured out that the values ranged from 0 to
943. Does "fa" denote something else and not fractional
anisotropy?
According to your paper published in Human Brain Mapping, "md" is
the average of the eigenvalues of the diffusion tensor and "l1" is
the largest of the eigenvalues. Is "lt" the average of the other
two eigenvalues, i.e., (l2+l3)/2 ?
Do the eigenvalues l1, l2 and l3 are the values in the x, y, and z
co-ordinates of a global Cartesian system, or every set of (l1,
l2, l3) corresponds to a local orthogonal system of the
eigenvectors of the diffusion tensor? If they correspond to a
local orthogonal system is it possible for me to get the
eigenvectors for each eigenvalue?