Your question makes sense but it could involve a bit of work to
realise the idea.
To summarise what is in the imported images:
* All the voxel values need to be zero or above. If a value is
unknown, then you can make it NaN.
* The values at each voxel in all the imported data of a particular
subject should sum to 1 or less.
* All image dimensions should be the same across imported data of all subjects.
Matrices in the header.
* The SFORM matrix, which is the one usually used for identifying the
orientation of the image, should be the same for all subjects. This
matrix is encoded by the srow_x, srow_y and srow_z fields of the NIfTI
* The QFORM matrix will probably differ across subjects. This matrix
is usually ignored by SPM, and is encoded by the pixdim, quatern_b,
quatern_c, quatern_d, qoffset_x, qoffset_y, qoffset_z and xyzt_units
fields of the NIfTI header. This matrix encodes the mapping to native
space, and is what allows Dartel deformations to work on native space
Other fields are largely ignored, except for the usual sizeof_hdr,
datatype, bitpix, vox_offset, qform_code, sform_code, scl_slope,
scl_inter & magic fields of the NIfTI file that SPM uses.
You may need to refer to some of the SPM code to actually generate
such images, although I guess you could take existing imported images
and modify them.
One approach that has been used to incorporate a bit more detail into
the imported data involves extending the new segmentation so that it
attempts to generate more different tissue types. This involves
generating additional tissue probability maps. Chris Lambert (my
former PhD student) did this with some success.
On 27 February 2012 13:15, Hikaru Takeuchi <[log in to unmask]> wrote:
> I would like to make dartel import image (usually represented as rc*) of certain images.
> I am trying to utilize DARTEL normalize procedures for normalization of FA images.
> However, when the normal DARTEL procedures are used (with a little twist to apply segmentation procedures to FA images), it turned out while outer edge of the white matter parts of the FA images are well aligned among different subjects, within the deep white matter areas, white matter places with different FA values are not aligned well. So, I would like to use images of the white matter parts of the FA map, instead of images of white matter segmentation (rc2*) in the DARTEL procedures, so that the information of difference of FA in the white matter tissue can be utilized in the normalization procedures.
> Can anyone advise on how to create dartel import image (r-) of arbitrary images? Or do the abovementioned dartel procedures with images of the white matter parts of the FA map make sense?
> Thanks in advance.