Dear Paolo,
I am not an expert, I'll just put forward some thoughts.
In my opinion, the size of the structures you want to analyze should be much greater than your native resolution (for example, 1-2 orders of magnitude). Your native voxels are about 70 cubic mm large, which is not far from the volume of the LGN in humans.
I have been analyzing thalamic nuclei by means of 3T fMRI (3mm x1.7mm x1.7 mm native resolution), and the structures I analyzed were 20 times greater than the native resolution of my voxels.
I think there are several ways to push on your resolution: you may acquire more slices at the cost of an increased TR; you may use a higher magnetic field (e.g. 7 T; see publications by
Metzger and Eckert); you may use an ad hoc normalization procedure (see what Prèvost et al. 2011 did for the amygdala); or you may use multivariate pattern analysis to make inferences on subvoxel resolution, but you would not be able to select only the LGN and the entire study would be different.
Hope it helps,
Giulio
--
Dear SPMers
I should perform an fMRI study to investigate a very small area such as the lateral geniculum body. Currently I acquire a voxel size of 3.75 x3.75x5 and normalize it into 2x2x2 voxel in spm. I'm worried about not being able to catch the signal from such a small area. I should adopt special procedures in the process of normalization?
Many thanks in advance
Paolo