Hello,
My name is Martin Zalesak, and I am an undergraduate Engineering
student at the University of Pennsylvania currently working on a
project in the Radiology Department of the University Hospital. The
aim of the project is to evaluate the utility of the SPM method for
clinical PET studies.
All of the applications that we would like to test essentially require
the comparison of a single patient single scan with an average image
of a pool of controls. My question is concerned with using the SPM
package for this purpose.
In more detail, the steps we would like to carry out are as follows:
1. Use the SPM package to realign, normalize and smooth
both control group images and the subject image.
2. Obtain an average image from the control group images.
This is a standard feature of the SPM package.
3. We would then like to obtain the variance and standard deviation
associated with each voxel of this average image.
4. We would then like to subtract the realigned, normalized and
smoothed image of the single subject study from the average image
of the control group. The result would be a "difference image"
(positive and negative). The we would like to see if the difference
image exceeds for example 3 standard deviations for that pixel
in the group image.
This could be done in the following way:
5. We would then want to set a threshold for each voxel, a multiple
of the standard deviation associated with that voxel in the group
average image obtained in step (3). This threshold for each of the
voxels would then define the "threshold image".
6. Finally we would want to display projection images of these
thresholded images to locate and identify areas of abnormality.
My questions are if there are ways of doing steps 3-6 using the SPM
package (we are currently using SPM95) and its standard routines.
Also, is it possible to convert the snrspm*.img matrices to a
format for manipulation in Matlab, and to feed any results back into
SPM to use their use their contiguous volume search, Tallairach
map locator and projection display capabilities?
Finally, failing the above is it possible to generate pseudo
group data from the single subject image by applying a random
number generator to the image with a small standard deviation and
using this pseudo group for analysis?
Thank you for your help,
Martin Zalesak
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|