Thanks for these explanations and the original post Pieter and and Michael. These are very enlightening.
I had a similar problem recently, not as many time points, (two conditions, two time points per condition) and we went with calculation of the perfusion maps in native space, then applying the transform of the M0 image to standard space (through concatenation of the transform of each time points structural image to standard space with the transform of structural to standard space). This worked well - but we had another problem. Our data was collected as 6 slices - and these did not always overlap consistently - that is, the top and bottom slices may not have had perfect alignment, and occasionally angulation of the slices was not the same. This meant that some perfusion maps (in standard space) extended below others (FH/IS direction), some times only into one part of the brain in standard space, while others may have extended above (both within participant time points and between participants). This of course would be a problem in any image based comparisons (say using randomise) across time points and conditions. To address this, I basically extracted the section ("slices") of the registered perfusion maps which did overlap (in the FH direction) in standard space to run our analysis.
If you are planning on using atlases to make ROI's for comparisons, it's likely not as important to be aware of this, or if you have "whole brain coverage" it is less of an issue, but if you are restricted to only a few slices in your acquisition, over lap of the slices between subjects is another factor you might want to consider - either by making a best effort to standardise prescription of the slices (what we will be improving on going forward), or in some fashion after the fact. Michael, do you have any guidance on the "best" practice for this issue (miss-match of slice overlap), or is there some where you could point me where it has been discussed? While what we did "worked" I'm more then happy to try a different method in the future if it it is considered better.
Thanks once again for the two interesting and useful posts.
Paul.
|