Hi all,
I'm using DARTEL for a VBM analysis in SPM5, comparing a patient and
control group. I'd like to take advantage of the ability to normalise
to MNI space, however in the guide the MNI spatial normalisation steps
are in a different section than the rest of the DARTEL steps, and I'm
having trouble telling where they'd fit in. Here's the order as I
understand it.
1) DICOM Import to get s*.img file
2) Display image and manually reorient to ACPC
3) Segment s*.img (GM and WM in native space; save bias corrected).
Produces bias corrected ms*.img
4) Segment the bias corrected image ms*.img (GM and WM in native
space; don't save bias corrected). Produces c1ms*_seg_sn.mat and
c2ms*_seg_sn.mat
5) Initial import - input all subjects' *_seg_sn.mat files. Produces
rc1ms*.nii and rc2ms*.nii
6) Create template - input all rc1ms*.nii and rc2ms*.nii. Produces
templates and flow fields (u_rc1ms*.nii)
7) Create warped - specify u_rc1ms*.nii, rc1ms*.nii, and rc2ms*.nii
files; select "modulation", 64 time steps, trilinear interpolation.
Produces mwrc1ms*.nii and mwrc2ms*.nii
8) Affine transform of DARTEL template to MNI space (1.2.1 in DARTEL
guide), using all options detailed in the guide
9) Combine deformations (1.2.2 in guide), using options detailed in
guide. Apply to mwrc1ms*.nii and mwrc2ms*.nii for each subject,
produces wmwrc1ms*.nii and wmwrc2ms*.nii
10) Smooth wmwrc1ms*.nii and wmwrc2ms*.nii at 8mm FWHM to produce
swmwrc1ms*.nii and swmwrc2ms*.nii
11) Run analysis on swmwrc1ms*.nii for GM or swmwrc2ms*.nii for WM;
two-sample t-test between patient and control, using absolute mask at
0.8.
I have segmentation run twice, once to produce the bias corrected
image and another to segment the bias corrected image. This comes
after reading John's response to the "unified segmentation and bias
correction - two steps?" question. If I have misinterpreted the
solution suggested by John, please let me know. And if I'm just
running the first segmentation to get the bias corrected image, can I
change the GM and WM output files from "Native Space" to "none" to
save time?
As I said up top, I'm not sure if the MNI spatial normalisation steps
(8 and 9 in the list above) are in the correct place, and therefore if
they're being applied to the correct images. Also, if the flow fields
are being applied to the images in Step 9 as part of the combined
deformation, should they still also be applied in step 7?
Hmmm, I realize that's a lot of questions/uncertainties... I hope I've
asked them relatively clearly. Thanks for any
help/clarification/confirmation you can provide. I like the DARTEL
procedure, and I expect I'll like it even more when I'm confident that
I'm doing it correctly!
Thanks,
Patrick Pruitt
|