I have some DWI/DTI images which were upsampled by the scanner from its
original 128x128 to 256x256. I would like to downsample the images back to
128x128 (I have read the 256x256 will cause many problems, will also require
significantly longer processing times, and of course is not as isotropic as
I originally intended). I have read that ApplyXFM can be used for this, but
do I need to have a transformation matrix previously created by FLIRT, or
can I create one myself? If the latter is possible how exactly do I go about
making one (details for this specific downsampling would be helpful)?
A second question is, a colleague of mine is presently writing a script to
downsample the DWI data, however, this will just be taking the center of the
256x256 k-space (the 128x128 portion) to downsample the image (I do not have
the phase information). Is this similar to what ApplyXFM would be doing? How
will removing the higher frequencies affect the calculation of eigenvectors
and eigenvalues?
Thanks,
Carlos Faraco
|