I am analyzing fMRI data with SPM 8 : one single subject, 4 sessions, aquired on two separate days.

I have noticed a systematical problem when checking results at p 0.99 : activations are far away from anatomical edges at different points, notably in the occipital poles, in anterior temporal lobes and in orbitofrontal cortex. See Im1.pdf attached.

Because such distorsions are not apparent in coregistered and normalized images, we figured it had something to do with the ffx analysis and, more specifically, with the default mask used by SPM at the step of model specification (See SPM_Mask.pdf).

So we applied a much more liberal explicit mask including all the normalized brain and even the skull. We did this by loading this liberal mask at the specification step in 'explicit masking'. Results are unchanged, and there is not one voxel activated outside of the brain.

Is it possible that such distorsions are only due to signal drop out in these regions? Strangely, even when asking SPM to use the explicit masking, it creates a default mask which edges correspond to the distorsions observed (just as in SPM_Mask.pdf). How can I make sure SPM really takes into account Explicit Masking and not the default masking?

If anyone has an idea on what is happening here I would very much appreciate any suggestion on this!

Thanks in advance,

Giulia

--