Dimension mismatch
lpxdjs
Posted on 08/05/10 11:05:34
Number of posts: 22
I have an issue that I'm having trouble trying to understand. I've looked through the forum but can't find any related topics, so any assistance would be greatly appreciated...
Before running my PLS analyses I registered all my pre-processed native image to a 4x4x4 MNI-152 standard space template. I've checked the header of the standard space image, as well as those of the registered native images, and they read as follows:
X orientation: R-L (i.e. radiological); 46 voxels; x=90 - x=-90mm
Y orientation: P-A; 55 voxels; y=-126 - y=90mm
Z orientation: Inf.-Sup.; 46 voxels; z=-72 - z=108mm
Image centre = anterior commissure
I would expect this header information to be preserved in the PLS results, which would allow me to choose my seed voxel location on the basis of activation peaks from an fMRI analysis conducted in FSL. Visual inspection of my PLS analysis output, however, shows the following:
X orientation: L-R (i.e. neurological); 46 voxels; x=-88 - x=92mm
Y orientation: P-A; 55 voxels; y=-104 - y=112mm
Z orientation: Inf.-Sup.; 46 voxels; z=-88 - z=92mm
Image centre = dorsal thalamus
Because of these discrepancies, the location of a peak voxel identified from my FSL analyses (x=-47, y=-68, z=8mm; voxels 137, 58, 80) lie outside of my PLS result image (same world-based coordinates are now located at voxels 11, 10, 25). Where might these discrepancies be coming from?
Thanks in advance,
Dan.
Untitled Post
I'm Online
jshen
Posted on 08/05/10 11:32:08
Number of posts: 291
No header information is changed.
If the images that you loaded to create datamat is in NIfTI format, the orientation and originator are determined by sform/qform in its header.
If they are in Analyze75 format, the orientation and originator are untouched, unless the originator is not specified. In this case, a point close to geometric center will be used for originator.
I cannot look into this further without your data. So please upload your data to ftp site and send me the link if you want me to further look into this issue.