Content-type: text/html
Computes tensor shape statistics based on the eigenvalues l1 >= l2 >= l3. The full list of statistics is
<cl> = (l1 - l2) / l1 , a measure of linearity,
<cp> = (l2 - l3) / l1 , a measure of planarity,
<cs> = l3 / l1 , a measure of isotropy.
Note that
cl + cp + cs = 1 .
<l1> = first eigenvalue
<l2> = second eigenvalue
<l3> = third eigenvalue
<tr> = l1 + l2 + l3
<md> = tr / 3
<rd> = (l2 + l3) / 2
<fa> = fractional anisotropy. (Basser et al, J Magn Reson B 1996)
<ra> = relative anisotropy (Basser et al, J Magn Reson B 1996)
<2dfa> = 2D FA of the two minor eigenvalues l2 and l3
sqrt(2 * [ (l2 - <l>)^2 + (l3 - <l>)^2 ] / (l2^2 + l3^2) )
<l> = (l2 + l3) / 2
The input to the program is the output of dteig.
Compute the linear statistic for a data set
dteig < A.inv1.Bdouble | dtshape -stat cl -outputdatatype float > A.inv1.cl.Bfloat
Compute RA for a data set, output as NIFTI:
cat dt_eig_sys.Bdouble | dtshape -stat ra -header data.nii.gz -outputfile ra.nii.gz
Compute the planar statistic cp for a data set with two tensors in each voxel
dteig -inputmodel multitensor -maxcomponents 2 < A.inv31.Bdouble | dtshape -stat cp > A.inv31.cp.Bdouble