Hi
I’ve decided to do the VBM on FA maps with the
following pipeline but still I have some points to decide:
- Eddy and motion correction. I
don’t have fieldmaps to make geometric distortion due to susceptibilities
changes. I don’t know a method to correct signal loses…
- Normalization of native images
to a standard space using linear affine 12 parameters. This can be done without
nonlinear warping to reflect variability in a common standard space, my
goal in this step is to make a diffusion template. The possibilities can
be:
- Native b0 ŕ EPI template
- Native whole head EPI (obtained
using the common sequence to functional BOLD imaging) (previously registered
to b0 native) ŕ EPI
template
“I think option a. quarantines
better registration of diffusion data to standard space. However, I propose b. because
I’m afraid b0 intensity is too low and maybe this can affect the
registration process in someway. Furthermore, although one uses a single shot
EPI sequence like functional BOLD imaging (which is the of the same modality of
EPI template) to obtain b0 image, some differences can be found between b0 and
BOLD-EPI that makes me suspect they are not exactly of the same modality (I
think, anyway, this can be circumvented using mutual information).
Finally, I can consider, as the
reference image or template, the T2 template if this entails better results
- I’d like to somehow mask
images to remove edge effects that arise when a wrong brain extraction is
done. This is very important because hyperintensities appear in the edges
of the FA maps due to the very noisy nature of diffusion images in voxels
outside the brain. In my opinion this is unacceptable for the construction
of a template.
- I can mask the b0 (again the
low intensity problem arises). This time I can use BET-FSL or some
masking routine in SPM I don’t know.
- I’ve tried probability segmentations
to define automatically an inside brain region (this leave most of the
time a fine line of hyperintensity)
- Once all natives images are in
standard space to obtain mean image. This will used as the reference image
to re-normalize the diffusion b0 images. In this step, I don’t know if
nonlinear warping should be used.
Is there any other step missing here (e.g. intensity
correction)?
Please, I’ll appreciate if a hard criticism if done.
Thank you for your time…
__________________________
Pedro Antonio Valdés Hernández
Neurophysics Department,
Cuban Neuroscience
Center.
Havana City, Cuba.
Phone Number: (537) 2084460
__________________________