Dear SPM gurus,
I've been going now through my first group analysis, and got the disturbing
result of being unable to obtain activations close to the most superior
points near the interhemispheric fissure.
What I did was:
realign (coregister & reslice) -> normalize (3mm voxels) -> smooth ->
statistical analysis (10 sessions, repeated conditions).
Looking more closely at my images I found that in each step I lose a little
bit off the top, presumably due to interpolation. I think that I can give up
the first reslicing and normalize the coregistered functionals,
circumventing the associated loss, and perhaps by writing the normalized
images at a higher resolution I can lose a little less there.
However, I noticed that the actual normalization process leads to some
variability (which I assume is natural) in the location of the brains of the
individual subjects along the z dimension. What is most disturbing is that I
was unable to obtain activations above the 'highest' brain voxel of the
'lowest' subject, as if an 'AND' masking operation takes place during the
statistical analysis. Physiologically I actually expected it to be higher.
Is this true, or is this the true location? If there is a masking taking
place, what, other than rejecting a couple of 'low' subjects, can be done?
What other suggestions can help me in this up-hill (roughly 5mm) battle?
thanks in advance,
Shy
|