 |
 |
Dear all, I am using SPM'99 to analyse animal (rat) pharmacological fMRI data. A groups of subjects were administered a compound of interest whilst in the scanner, after a certain number of baseline scans. The data is analysed using the "multi subjects with covariates" in PET module (because we donšt use EPI, our MRI scans are quite long, >1 min per timepoint, and better resemble PET scans in terms of temporal resolution). Covariate of interest i.e. input function was a curve derived from the drug's pharmacodynamics. This is a simple curve which increases from baseline, peaks and then slowly drops back to near baseline over time. I also included realignment parameters as nuisance variables. All brains were normalised, smoothed etc. This kind of procedure has been successfully applied to animal's fMRI data by several colleagues in our group. A specific problem arises- I have included global normalisation with ANCOVA (having checked that global brain signal is not significantly correlated with my paradigm of interest, taking into account warnings by Aguirre at al, 1998). SPM's of positive correlation with input function (choosing 1 as a contrast of interest) revealed a number of positively activated regions - red blobs. However, the timecourse of one such activated cluster derived by Marsbar (choosing scaling form spm.mat) does not appear to relate positively to the input function. In this timecourse the signal decreases from baseline over time; given the positive significant correlation with the input function one would expect signal to increase. Timecourse derived using VOI (non-adjusted option) reveals an almost identical timecourse. However, choosing the adjusted VOI (for effect of interest) gives me a much noisier timecourse which vaguely behaves in the "correct way" (increases rather than decreases). I should also add that the global brain signal itself decreases over time, showing what is probably a physiological drift. I am aware that this decrease in global signal is likely to be the main influence upon the non-adjusted timecourse. Based on the above, however, I have few questions: 1) Regarding Marsbar: I expected Marsbar timecourse to take into account (and remove the effect of) global brain signal since this was scaling applied in the analysis. I selected the option to include scaling from spm.mat during Marsbar calculation. Is anyone familiar with why this has not happened or is there another explanation? 2) About VOI: does the adjusted timecourse represent the signal timecourse with global signal component removed or does it include much further processing? 3) But the main question is: how can I display the timecourse that spm 'sees', in other words the one that matches the activation map? Thanks in advance, Diana Cash