> How does one go about constructing a T2* template that is in good
> correspondence with the T1 template? How did you create EPI.img? We are
> interested in doing something similar with our spiral T2* images.
The basic procedure for making a T2* template was:
1) Estimate the affine normalization parameters that get a number of
T2* images to the same stereotactic space.
2) Write the spatially normalized images with 2mm isotropic voxels and
a bounding box of -90..91, -126..91, -72..109.
3) Generate a "soft" mean of the images (to use Roger Wood's terminology).
4) Smooth this mean image with an 8mm FWHM Gaussian.
Step 1 is the tricky one. My original attempt at estimating spatial
normalization parameters involved co-registering a T1 weighted structural
image to the functional images, and determining the affine normalization
parameters from the T1 by matching it to a template. However, the T2*
(EPI) template at ftp://ftp.fil.ion.ucl.ac.uk/spm/data/EPI_template.tar.gz
was generated a different way. I segmented the images to extract the gray
matter, and then spatially normalized the gray matter to the gray matter
probability image. The segmentation begins with an affine registration
step, so I used the EPI template that I created with my first attempt.
Generating a "soft" mean can be done using the ImCalc button. Just select
the input images, enter the name of an output image and give it a function
such as:
(i1+i2+i3+i4+i5)./((i1~=0)+(i2~=0)+(i3~=0)+(i4~=0)+(i5~=0))
This example will create the soft mean of five images.
I would guess that the smoothing should really be done on the images
before they are spatially transformed, but I don't think that doing it
afterwards should make too much difference.
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|